2,673. published advisories
ShieldMap · AI security, mapped
How fast the problem is growing.
- Packages and products
- 295
- Advisories
- 2,924
- Malicious packages
- 161
- Exploited in the wild
- 13
Advisories by the month they were published, since MCP was released in November 2024. The busiest month so far was Mar 2026, with 366. A red dot marks a package with an advisory on CISA's exploited list.
Nov 202659 advisories · 25 malicious packages
- HighJupyter MCP Server: Unauthenticated management routes can hijack the Jupyter backend in standalone HTTP modejupyter-mcp-server
- CriticalJupyter MCP Server: Browser DNS rebinding can hijack the Jupyter backend through management routesjupyter-mcp-server
- CriticalMalware in @guicai777/mcp-doctor@guicai777/mcp-doctor
- CriticalMalicious code in @guicai777/mcp-doctor (npm)@guicai777/mcp-doctor
- Highmssql-mcp-core: SQL injection in mssql-mcp write tools via unparameterized table names and WHERE clauses@connorbritain/mssql-mcp-core, @connorbritain/mssql-mcp-reader, @connorbritain/mssql-mcp-server, @connorbritain/mssql-mcp-writer
- MediumSafari MCP: JavaScript injection through single-quoted tool arguments (backslash not escaped)safari-mcp
- CriticalTodoist MCP Server: Browser DNS rebinding lets websites write and delete Todoist data through local MCP server@doist/todoist-mcp
- CriticalMalicious code in @fdcasj_dfew/md-doc-mcp (npm)@fdcasj_dfew/md-doc-mcp, @fdcasj_dfew/md-doc-mcp
- Highk8s-mcp-server: Missing authentication on default SSE/HTTP transport allows unauthenticated Kubernetes tool executiongithub.com/reza-gholizade/k8s-mcp-server
- Highobsidian-sync-mcp: Missing Host/Origin validation lets a public website drive a victim's Obsidian vault via DNS rebinding (no-auth mode)obsidian-sync-mcp
- HighOpenClaw delegation could skip approval for restricted agentsopenclaw, openclaw
- Highsqlite-mcp: Read-only mode bypass via statement chaining and SQL injection in describe_tablegithub.com/StacklokLabs/sqlite-mcp
- HighModel Context Protocol Python SDK: OAuth client could send credentials to an authorization server chosen by the MCP servermcp, mcp
- HighAxon: Download of Code Without Integrity Check in axon-mcpaxon-mcp
- Medium/v1/systemone bypasses the --trust-request-chat-template gatevllm
- CriticalMalware in @ifasn/ms-doc-mcp@ifasn/ms-doc-mcp
- CriticalMCP Path Traversal Allows Full-Filesystem Read and JWT Secret Exfiltration Leading to Account Takeoverlangflow, langflow-base
- CriticalMalicious code in mcp-aevo-server (PyPI)mcp-aevo-server, mcp-aevo-server, mcp-aevo-server, mcp-aevo-server, mcp-aevo-server, mcp-aevo-server
- CriticalPhoenix: Format String Injection in applyChatTemplatearize-phoenix
- CriticalMalware in vamp-mcpvamp-mcp, vamp-mcp, vamp-mcp
- CriticalMalware in @ifasn/mmdd-doc-mcp@ifasn/mmdd-doc-mcp
- CriticalMalware in @shadowdu/memos-mcp@shadowdu/memos-mcp
- CriticalMalware in @shadowdqj/image-mcp@shadowdqj/image-mcp
- CriticalMalware in mem0-mcp-servermem0-mcp-server
- Highcode-review-mcp-server: Command injection via unsanitized folderPath/baseBranch parameterscode-review-mcp-server
- CriticalMalware in karma365-mcpkarma365-mcp
- MediumIntegration Nodes Send Authenticated Requests to Unintended API Endpointsn8n, n8n, n8n, n8n
- MediumAWS Bedrock Nodes Skip SSRF Protection for Custom Runtime Endpointsn8n, n8n, n8n
- MediumWorkflow Activation Allows Trigger Nodes to Use Credentials the Publisher Cannot Accessn8n, n8n, n8n, n8n
- HighInvitation Acceptance Endpoint Lets a Member Take Over Pending User Accountsn8n, n8n
- HighServer-Side Request Forgery in the MLflow tracking server via client-supplied artifact locations and model version sourcesmlflow
- HighCustom Instance Role with User Management Permissions Gives Control of the Owner Accountn8n, n8n, n8n
- CriticalMalicious code in my-mcp-server-driving-license-ocr (PyPI)my-mcp-server-driving-license-ocr
- CriticalMalicious code in @rfewqa-fes/mnm-doc-mcp (npm)@rfewqa-fes/mnm-doc-mcp
- CriticalMalware in @go-gost/mcp-darwin-x64@go-gost/mcp-darwin-x64
- HighAG2: Newline injection in SandboxShellTool bypasses allowed=[...] and readonly=True, enabling arbitrary command executionag2
- CriticalMalware in mcp-contract-testermcp-contract-tester
- CriticalMalware in @lemili9804/mcp-weather-server@lemili9804/mcp-weather-server, @lemili9804/mcp-weather-server
- HightsParticles: Unbounded HTTP Request Body Buffering in @tsparticles/mcp-server Leads to Remote DoS@tsparticles/mcp-server
- HightsParticles: Malformed JSON requests can terminate the @tsparticles/mcp-server HTTP process@tsparticles/mcp-server
- HighClaude Code: TOCTOU Race in Edit Tool Allows File Writes Outside the Workspace@anthropic-ai/claude-code
- Mediumfast-agent: `fast-agent serve` binds HTTP to 0.0.0.0 with no authentication by default; with --shell this is unauthenticated remote command fast-agent-mcp
- MediumThinking-budget sparse state aliases onto a co-batched request during a mixed `SWAP`, forcing the reasoning-end token into another request'svllm
- MediumtsParticles: A single anonymous MCP session can monopolize HTTP mode and deny new clients@tsparticles/mcp-server
- MediumCross-user token data leakage in async scheduling + Pipeline Parallelismvllm
- CriticalMalware in @bestcoolman/dev-mcp@bestcoolman/dev-mcp
- CriticalMalicious code in @massnerder/x402-scraper-mcp (npm)@massnerder/x402-scraper-mcp
- MediumResponse amplification in OpenAI validation error handlervllm
- MediumLangflow URL component: unbounded response size allows denial of service via memory exhaustionlangflow, langflow-base
- CriticalMalicious code in teamviewer-mcp-server (PyPI)teamviewer-mcp-server
- CriticalMalicious code in mcp-teamviewer (PyPI)mcp-teamviewer
- CriticalMalicious code in @ifasn/ms-doc-mcp (npm)@ifasn/ms-doc-mcp, @ifasn/ms-doc-mcp
- CriticalMalicious code in @ifasn/mmdd-doc-mcp (npm)@ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp, @ifasn/mmdd-doc-mcp
- CriticalMalicious code in @ifasn/mdd-doc-mcp (npm)@ifasn/mdd-doc-mcp, @ifasn/mdd-doc-mcp, @ifasn/mdd-doc-mcp
- HighMissing request-wide media budget allows single-request frontend resource exhaustionvllm
- CriticalMalicious code in @ifasn/mmd-doc-mcp (npm)@ifasn/mmd-doc-mcp, @ifasn/mmd-doc-mcp
- HighUnauthenticated EngineCore crash via out-of-range prompt token id under skip_tokenizer_init (incomplete fix of GHSA-25q3-v2hm-8vpf)vllm
- MediumvLLM Speculators Config verifier.name_or_path Element Directory Traversal Information Disclosure Vulnerabilityvllm
- CriticalMalicious code in @ifasn/md-doc-mcp (npm)@ifasn/md-doc-mcp, @ifasn/md-doc-mcp, @ifasn/md-doc-mcp
Oct 2026175 advisories · 2 malicious packages
- HighPraisonAI vulnerable to Remote Code Execution via Broken AST Sandboxpraisonai
- High@sylphx/pdf-reader-mcp: HTTP transport accepts unauthenticated requests despite a configured MCP_API_KEY@sylphx/pdf-reader-mcp
- MediumSQL Injection via Metadata Filter Key in the Vector Store (OracleDB) Noden8n, n8n, n8n
- MediumLangGraph: Unencoded identifiers in LangGraph SDK v3 streaming can redirect authenticated GET requestslanggraph-sdk, langgraph
- MediumOAuth Account Hijack in the End-User Credential Connect Flow via Unauthenticated Callbackn8n, n8n, n8n
- MediumMissing authorization on gateway secret creation in MLflow basic-auth with workspaces enabledmlflow
- MediumONNX: Out-of-Bounds Read in RNN/GRU/LSTM Shape Inference (Old Opsets, Missing Rank Check)onnx
- HighLangGraph JavaScript custom authentication may permit cross-user Store accesslanggraph-api
- Highmcp-sanitizer: URL sanitizer allows localhost/IPv6 loopback SSRF bypass via explicit port or bracketed IPv6mcp-sanitizer
- MediumMongoDB Node Update Operations Can Modify Unintended Documentsn8n, n8n, n8n, n8n
- MediumMem0: Cross-tenant conversation leak via unescaped delimiter collision in session-scope key (_build_session_scope)mem0ai
- MediumvLLM: DoS via unbounded chat_template Jinja2 rendering exhausts the renderer thread poolvllm
- CriticalMalware in @atlisp/mcp@atlisp/mcp
- HighvLLM: Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabledvllm
- MediumONNX: Out-of-bounds read in ConvTranspose shape inference via undersized weight rankonnx
- Highmcp-sanitizer: File Path Sanitizer Permits Arbitrary User-Home Absolute Pathsmcp-sanitizer
- MediumvLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM enginevllm
- MediumLangGraph CLI prints the registry push token in `langgraph deploy --verbose` outputlanggraph-cli
- HighDeprecated Code Nodes Allow Code Execution in the n8n Main Processn8n, n8n, n8n
- MediumBrowser Use Upload Tool Reads Arbitrary Files from the n8n Servern8n, n8n, n8n
- HighInsecure GraphicsMagick Configurationn8n, n8n, n8n, n8n
- MediumvLLM DeepSeek-OCR image processing vulnerable to denial of servicevllm
- HighInsufficient Isolation in Workflow Expression Evaluationn8n, n8n, n8n, n8n
- HighvLLM: Untrusted EPS inputs can trigger denial of service or Ghostscript RCEvllm
- MediumNocoDB Node Sends Its API Token to Attacker-Chosen Hosts When Downloading Attachmentsn8n, n8n, n8n, n8n
- MediumONNX: Null Pointer Dereference in ScanInferenceFunction Due to Missing Required Attributeonnx
- MediumAsana Trigger: Unauthenticated Request Can Reset Webhook Verification Secretn8n, n8n
- MediumCredential-Using Nodes Forward Authentication Headers on Cross-Origin Redirectsn8n, n8n, n8n
- MediumChat Trigger Embedded Chat Does Not Enforce n8n User Auth on All Requestsn8n, n8n, n8n, n8n
- MediumONNX: Denial of service via unchecked expansion of packed sub-byte tensors in ONNX Python APIsonnx
- MediumvLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)vllm
- MediumOpen Redirect via MCP OAuth Dynamic Client Registrationn8n, n8n, n8n, n8n
- LowvLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP servervllm
- MediumvLLM: Responses API bypasses trust_request_chat_template for chat_template_kwargsvllm
- CriticalMLflow vulnerable to remote code execution on unauthenticated servers via third-party scorer deserialization importing caller-placed artifacmlflow
- HighStrands Agents: Local Privilege Escalation via Symlink Attack in Default Session File Storagestrands-agents
- MediumModel Context Protocol Python SDK: HTTP client transports followed cross-origin redirects carrying custom headers and bodiesmcp, mcp
- MediumvLLM denial of service via prompt embeds on XD-RoPE models (incomplete fix of CVE-2026-55514)vllm
- MediumStrands Agents: Default FileSessionManager storage is world-readable on shared systemsstrands-agents
- HighLiteLLM: Privilege Escalation to Proxy Admin via Cross-Domain Reuse of the Salt Keylitellm, litellm, litellm, litellm, litellm
- MediumvLLM OpenCVDynamicOpenPanguVideoBackend: request-level num_frames without a ceiling — unauthenticated memory/CPU exhaustion (OOM-kill) on anvllm
- MediumQuadratic Cohere2 Vision preprocessing DoS via request-controlled max_patchesvllm
- HighMCP TypeScript SDK: HTTP server transports and handlers did not limit request body size or JSON-RPC batch length@modelcontextprotocol/hono, @modelcontextprotocol/node, @modelcontextprotocol/sdk, @modelcontextprotocol/server
- Mediumvllm p2p never connected zmq_exhaustionvllm
- HighLiteLLM: internal_user_viewer arbitrary local file read via `vertex_ai_credentials`litellm
- HighvLLM: Request-controlled mm_processor_kwargs.code_revision allows remote code executionvllm
- MediumLiteLLM vulnerable to reflected XSS in /sso/debug/callbacklitellm
- MediumMLflow CrewAI autologging records LLM provider API keys in logged trace spansmlflow
- CriticalAgno: Arbitrary code execution via `eval()` on `field_type` in `FunctionCall`agno
- HighMLflow vulnerable to arbitrary file read on unauthenticated servers via prompt model version sourcemlflow
- Mediumsanitize_message() address regex is lowercase-only and does not match PIL's uppercase __repr__, leaking live heap addresses to unauthenticatvllm
- MediumModel Context Protocol Python SDK: Client fetched server-chosen $ref URLs while validating tool results against output schemasmcp, mcp
- MediumIncomplete CVE-2026-34755 remediation can be bypassed by request-level num_framesvllm
- MediumMiniMax M3 video preprocessing permits request-driven CPU resource exhaustion through min_pixelsvllm
- MediumModel Context Protocol Python SDK: Server bearer-token authentication accepts tokens issued for a different resource servermcp, mcp
- High@justice8096/sagemath-mcp-server: executes arbitrary code via the escapeShellString and sanitizeCode functions@justice8096/sagemath-mcp-server
- MediumMLflow: SSRF on unauthenticated MLflow servers via third-party scorer kwargs in the scorer invoke, register, and GenAI evaluate endpointsmlflow
- HighStrands Agents SDK: Exposure of data element to wrong session in the A2A Serverstrands-agents
- MediumLangGraph: Namespace labels containing `.` reach other namespaces in Postgres and SQLite storeslanggraph-checkpoint-postgres, langgraph-checkpoint-sqlite
- MediumFastMCP: Post-transform component authorization is skipped on direct callsfastmcp, fastmcp, fastmcp-slim, fastmcp-slim
- High@rog0x/mcp-docker-tools: executes arbitrary code via the handleImageList function@rog0x/mcp-docker-tools
- HighvLLM : DoS via unvalidated mm_processor_kwargs in multimodal image processors (Isaac, InternVL, H2OVL, etc.)vllm
- High@rog0x/mcp-git-tools executes arbitrary code via the resolveBaseBranch function@rog0x/mcp-git-tools
- MediumvLLM: Forged multimodal UUID + P0/P1 processor-cache eviction drift: unauthenticated remote crash of the whole vLLM enginevllm
- MediumFastMCP MCP client wrappers construct M2M OAuth providers without issuer binding — client secret redirectable to attacker-advertised authorifastmcp, fastmcp-slim
- HighJupyter Server: Gateway WebSocket handler bypasses kernel execute authorizationjupyter-server
- Criticalmem0ai Databricks vector store vulnerable to SQL injectionmem0ai
- Highnbconvert: Arbitrary file write through path traversal in cell output filenames when exporting to md, latex, rst and asciidocnbconvert
- HighStorybook dev server vulnerable to code injection from the local network@storybook/addon-mcp, @storybook/addon-mcp
- MediumOpenClaw: Browser CDP discovery could accept blocked WebSocket URLsopenclaw
- CriticalStorybook: Missing origin validation in Storybook preview allows code injection from malicious websites@storybook/addon-mcp, @storybook/addon-mcp
- HighMCP TypeScript SDK: Experimental tasks were not tied to the session that created them@modelcontextprotocol/sdk
- HighMCP TypeScript SDK: One HTTP request within the default size limit could exhaust server memory during validation@modelcontextprotocol/sdk, @modelcontextprotocol/server
- MediumMCP TypeScript client SDK: HTTP client transports and OAuth requests followed cross-origin redirects and re-sent custom headers and request @modelcontextprotocol/client, @modelcontextprotocol/sdk
- HighMCP TypeScript server SDK: Shared transport or server instances delivered one HTTP client's messages to another@modelcontextprotocol/server
- MediumMCP TypeScript SDK: Server bearer authentication accepted tokens issued for another service@modelcontextprotocol/express, @modelcontextprotocol/sdk, @modelcontextprotocol/server, @modelcontextprotocol/server-legacy
- HighMLflow FileStore trace IDs allow path traversal and arbitrary file writesmlflow
- MediumChannel members can post messages shown as a webhook via client-supplied message metadataopen-webui
- CriticalClaude Agent SDK: Argument Injection via resume Option Allows Arbitrary Command Executionclaude-agent-sdk
- Lowckan-mcp-server: DataStore Table UI resource does not validate postMessage origin (in) and posts to targetOrigin "*" (out)@aborruso/ckan-mcp-server
- MediumMCP Python SDK: Unbounded Session Accumulation in StreamableHTTP Transport Enables Resource Exhaustionmcp, mcp
- MediumJupyterLab: Command link trust prompt in JupyterLab can trust a document other than the one clickedjupyterlab, jupyterlab, notebook
- HighClaude Code: Plugin Hash Pin Bypass Allows Bypass of Hash Pinning@anthropic-ai/claude-code
- MediumUnauthenticated Slack Token Exfiltration and Full-Response SSRF via the Slack Trigger Webhookn8n, n8n, n8n
- MediumChat Hub: chat-only users can register tool types the product already tries to sandboxn8n, n8n, n8n
- MediumQuick Base Node Allows Filter Injection Through the Where Parametern8n, n8n, n8n
- MediumExecute Sub-workflow `source: url` / `source: localFile` Bypasses Shared-Workflow Credential Guardn8n, n8n, n8n
- MediumCredential Permission Check Skips Inline Sub-Workflow Nodes With a Non-Boolean Disabled Valuen8n, n8n, n8n
- MediumSeaTable Trigger Node Stores the API Token in Failed Poll Execution Datan8n, n8n, n8n
- MediumLocal Gateway Deep-Link Connect Silently Re-pairs to Attacker-Controlled Instancen8n, n8n
- HighPrototype Pollution in Workflow History Compaction Causes Denial of Servicen8n
- HighRead/Write Files Node Glob Pattern Causes Unbounded Event-Loop Blockn8n, n8n, n8n
- MediumOData $filter Injection in Microsoft Outlook and SharePoint v1 Nodesn8n, n8n, n8n
- HighWebhook Responses Without the Sandbox Policy Allow Stored Cross-Site Scriptingn8n, n8n, n8n
- MediumPer-User Credential Theft via Error-Workflow Execution Context Inheritancen8n, n8n
- MediumCredential Secrets Exposed in Execution Error Data of Several Nodesn8n, n8n, n8n
- MediumCredential Permission Check Skips Nodes Whose Default Parameters Are Omittedn8n, n8n, n8n
- HighCrafted Workflow ID Gives Access to Data Tables in Other Projectsn8n, n8n
- CriticalMalware in @revizahoshii/hoshino-mcp-server@revizahoshii/hoshino-mcp-server
- MediumRemote DoS in disaggregated multimodal /inference/v1/generate via malicious kwargs_data / mm_placeholdersvllm
- HighFrame-count cap bypass in VideoMediaIO.load_base64 for video/jpeg via media_io_kwargs.video.num_framesvllm
- MediumRust gRPC output_candidates.token_ids bypasses logprob_token_ids limits and can terminate EngineCorevllm
- HighRemote DoS in vLLM via allowed_token_ids Speculative-Decoding Metadata Mismatchvllm
- HighNemotron-VL processors disable Pillow decompression-bomb guard globally, enabling unauthenticated DoSvllm
- MediumMultimodal chat audio decoding bypasses the decompression-bomb duration limitvllm
- HighCross-tenant DoS via uninitialized LogprobsTensors with prompt_logprobsvllm
- MediumRust gRPC CandidateTokens.top_n bypasses max_logprobs and can terminate EngineCorevllm
- HighSSRF via X-Request-Id header drives P2pNcclConnector outbound ZMQ+NCCL pair, leaking KV-cache tensorsvllm
- MediumRust gRPC Generate accepts out-of-vocabulary prompt token IDs and can terminate EngineCorevllm
- MediumUnvalidated stop_token_ids causes engine crash (DoS) and out-of-bounds GPU writevllm
- MediumPraisonAI: Human-in-the-loop tool approval is cached by tool name and silently reused for all subsequent calls with arbitrary argumentspraisonaiagents
- HighPraisonAI: Call API localhost-only authentication bypass via spoofed Host headerpraisonai
- CriticalPraisonAI: CodeAgent Executes LLM-Generated Code Without Sandboxing and Leaks All Environment Secretspraisonaiagents
- LowPydantic AI OpenTelemetry instrumentation: retry prompt content is not redacted when `include_content=False`pydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- HighPraisonAI: Unsafe Dynamic Module Loading Leads to Arbitrary Code Execution via tools.py in AgentFlowpraisonaiagents
- MediumPraisonAI: Prompt-injection defense blocks only when 3+ detector families fire simultaneously; realistic single-vector injections pass throupraisonai
- HighPraisonAI: Jobs API is unauthenticated by default and allows attacker-controlled webhook SSRFpraisonai
- LowPydantic AI OpenTelemetry instrumentation: exception events on tool and agent run spans include content when `include_content=False`pydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- MediumPraisonAI: FastContext path resolution permits absolute and traversal reads outside the workspacepraisonaiagents
- HighPraisonAI: SecurityPolicy command/path/import restrictions are completely unenforced by the default SubprocessSandbox backendpraisonai
- HighPydantic AI: Concurrency-limited models can keep their slot when a streamed request ends earlypydantic-ai, pydantic-ai-slim
- HighPraisonAI: API deploy code generator embeds unescaped YAML fields into Python sourcepraisonai
- CriticalPraisonAI: AICoder Arbitrary File Write and Command Execution via LLM Tool Callspraisonai
- HighPraisonAI: SkillTools Executes Scripts Without Path Containment Validationpraisonaiagents
- HighPraisonAI: Shell command allowlist bypass via find -exec built-in actionpraisonai
- MediumPydantic AI: Event loop blocked by quadratic title extraction in `web_fetch`pydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- HighPydantic AI Web chat UI (`Agent.to_web()`, `clai web`): a website visited by the developer can trigger agent runs and server-side tool execupydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- HighPraisonAI: MCP HTTP-stream transport is unauthenticated by default, exposing tool enumeration and an unvalidated tool-call surfacepraisonai
- HighPraisonAI: Plugin Auto-Discovery Executes Arbitrary Python Files Without Verificationpraisonaiagents
- MediumPydantic AI Web chat UI (`Agent.to_web()`, `clai web`): the local chat endpoint does not validate the `Host` headerpydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- MediumPraisonAI: ContextGatherer include resolution permits absolute and traversal reads outside the workspacepraisonai
- HighPraisonAI: DNS rebinding bypass in `web_crawl` SSRF protection allows internal response disclosurepraisonaiagents
- MediumPraisonAI: Project config can auto-save agent output outside the project rootpraisonaiagents
- MediumPydantic AI: Unbounded memory use when downloading remote content via web_fetch or FileUrlpydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- MediumPydantic AI: Excessive resource use when local web fetching converts nested HTMLpydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- MediumPydantic AI: SSRF cloud-metadata blocklist bypass via IPv6 zone identifierspydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- MediumPraisonAI: PGVector and Cassandra knowledge stores interpolate vector dimensions into DDLpraisonai
- MediumPraisonAI: Project custom command templates can read outside-workspace files into model promptspraisonai
- HighLangflow: IP Spoofing Bypass via `X-Forwarded-For` Allowing Remote Configuration Writelangflow
- HighLangflow has Authenticated Cross-Project File Disclosure via Unscoped MCP Resource Handlerslangflow
- HighPraisonAI: Crawl4AI/Chromium backend is also affected by the `web_crawl` SSRF validation bypasspraisonaiagents
- HighPraisonAI: AgentMail webhook mode accepts forged unsigned message.received events and invokes agentspraisonai
- CriticalFlowise NodeVM sandbox escape via puppeteer allowlist - authenticated RCE and arbitrary file read via Chromiumflowise, flowise-components
- HighPraisonAI: Prompt Injection Defense Bypassed for HIGH-Level Threatspraisonai
- CriticalPraisonAI: Code Injection via f-string Interpolation in Deploy API Server Generationpraisonai
- MediumLangflow : Cross-user flow access and vertex execution via deprecated /api/v1/build/{flow_id}/vertices endpointslangflow, langflow-base
- HighPraisonAI: AgentMail webhook lacks signature verification, allowing unauthenticated message injection and sender spoofingpraisonai
- CriticalLangflow: OS command injection (RCE) via arbitrary command in MCP stdio server configurationlangflow, langflow-base
- CriticalFlowise Prompt Injection to RCE and SSRF via CSV/Airtable Agent Python Validator Bypassflowise, flowise-components
- HighPayload: Improper access control for MCP API keys@payloadcms/plugin-mcp
- MediumAn oversized min_tokens in one request hangs the vLLM engine while /health stays greenvllm
- LowV1 InputBatch.condense leaves a stale allowed_token_ids mask on recycled batch rowsvllm
- HighMCP TypeScript SDK: OAuth client could send credentials to an authorization server chosen by the MCP server@modelcontextprotocol/client, @modelcontextprotocol/sdk
- HighMOSS-Audio request-controlled processor cache permits remote memory exhaustionvllm
- CriticalLangflow: PythonREPLComponent executes unsandboxed Python code, enabling authenticated RCE and privilege escalationlangflow
- LowvLLM: Harmony tool continuations drop `cache_salt` — restoring a cross-tenant prefix-cache membership oraclevllm
- MediumLangflow: SSRF in URL-taking components (protection disabled by default / warn-only, not applied to RSS, SearXNG, Web Search, Home Assistantlangflow, langflow-base
- MediumvLLM: Mirrored multimodal IPC caches desync after a rejected request — a later request reusing the same media hash trips a receiver assertiovllm
- MediumCaller-controllable values become or collide with shared cache keys, enabling cross-request cache aliasing (2 sites: multimodal EXIF hash anvllm
- MediumvLLM: Loose `cache_salt` validation lets a single request kill EngineCore on LMCache-MP deployments — uncaught downstream `ValueError` deniavllm
- MediumvLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and inducevllm
- MediumvLLM: GLMGA video sampling permits request-driven CPU and memory exhaustionvllm
- MediumvLLM: Structured-output request errors escape the request boundary and terminate the shared EngineCore — engine-fatal denial of service (3 svllm
- HighLangflow: Prompt injection in Langflow Smart Transform can lead to code executionlangflow
- CriticalLangflow: Unauthenticated Flow Execution via Webhook Authentication Bypasslangflow
- HighLangGraph SDK custom auth silently ignores actions= on resource decoratorslanggraph-sdk
- MediumOpenClaw: Channel read actions could skip target allowlists@openclaw/feishu, @openclaw/googlechat, @openclaw/matrix, @openclaw/msteams
- CriticalLangflow: Weak Fernet Key via random.seed()langflow
- MediumvLLM: Scale-out disaggregated multimodal transport trusts caller-supplied featuresvllm
- HighLangflow: Title Authenticated Remote Code Execution in validate_code via Malicious Decorators Descriptionlangflow
- MediumvLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reachvllm
- Mediumrmcp OAuth client fetches server-controlled resource_metadata URLsrmcp
- MediumJupyterLab: Argument injection in JupyterLab extension uninstall exposes server-readable files and internal URLsjupyterlab, jupyterlab
- MediumJupyterLab: Cross-site scripting (XSS) in JupyterLab via crafted language package (jupyterlab.json)jupyterlab, jupyterlab
- HighJupyterLab: Cross-site scripting (XSS) in JupyterLab via notebook cells pasted from the system clipboardjupyterlab, jupyterlab, notebook
Sept 2026257 advisories · 20 malicious packages
- HighStored XSS via Same-Origin Blob URL in the Binary-Data File Previewn8n, n8n, n8n
- MediumCross-Project Agent Ownership Transfer via Client-Supplied Agent IDn8n, n8n
- MediumLiteLLM: Authenticated SSRF and provider-credential exfiltration via unvalidated request-body routing parameterslitellm, litellm, litellm, litellm, litellm, litellm, litellm, litellm, litellm
- MediumPrototype Pollution in the AI Workflow Builder Connection Mergen8n, n8n, n8n
- HighUnauthenticated Unbounded OAuth Client Persistence via the Authorize Endpointn8n, n8n
- MediumUnauthenticated Cross-Project Workflow Execution via Webhook Path-Only Resolutionn8n, n8n, n8n
- HighShared Prototype Mutation Through the MCP Workflow-Validation Interpreter Allows Owner Account Takeovern8n, n8n
- HighSQL Injection in the Microsoft SQL Node via Expression Interpolation into the Query Fieldn8n, n8n
- HighSend-and-Wait HMAC Bypass Allows Unauthenticated Approval of Waiting Executionsn8n, n8n
- HighExecute Sub-workflow Inline JSON Allows Member to Spoof Workflow Identity for Credential Authorization, Static Data, and Error-Workflow Dispn8n
- HighHTTP server transports and OAuth endpoints read request bodies with no size limitmcp, mcp
- MediumCross-User Agent Chat Resume Allows Hijacking Another User's Pending Tool Approvaln8n, n8n
- HighShared-Workflow Credential Check Misses Nested and Tool Inline Sub-Workflowsn8n
- HighShared-Workflow Credential Check Misses Agent Node Parameter Credentialsn8n, n8n
- Highn8n Chat Trigger Stored XSS via customCss Parameter on Hosted-Chat Pagen8n, n8n, n8n
- HighCode Execution in the Git Node Log Operation via Unneutralized Repository Configurationn8n, n8n, n8n
- LowLangflow 1.0.16 - 1.11.6: unsafe eval() on Tool-Mode input options in schema.pylangflow, langflow-base
- MediumCrafted IAMF audio upload reaches a PyAV/FFmpeg native heap overflow through the speech transcription path — denial of servicevllm
- HighTerminal users can manage Terminals policies and other users' terminals via the terminal proxyopen-webui
- MediumTool source code disclosed to read-only users via the tool export endpointopen-webui
- MediumRevoked users keep signing in via token exchange when roles are only in the ID tokenopen-webui
- MediumInstance-wide stall via recurrence rules with no possible occurrenceopen-webui
- MediumRead-only shared folder members can still add chats to the folder by forkingopen-webui
- MediumDeactivated users keep terminal command execution via an already open terminal WebSocketopen-webui
- MediumKnowledge base collaborators keep read and delete access to removed files via stale file metadataopen-webui
- MediumStored XSS via user-authored tool-call embeds in the Markdown rendereropen-webui
- HighAny authenticated user can steal another user's session token via script embedded in a DOCX previewopen-webui
- MediumAny authenticated user can repeatedly stall a worker via file-search regex on the event loopopen-webui
- MediumOther users' session tokens disclosed via OAuth-mode connections, tool servers and terminalsopen-webui
- MediumAny authenticated user can exhaust worker memory via a nested-repetition knowledge-search patternopen-webui
- MediumAny authenticated user can reach disabled model provider backends via an explicit backend indexopen-webui
- MediumDirectory structure of inaccessible knowledge bases disclosed via an unscoped directory idopen-webui
- HighViewer session token theft via a script link in a shared chat's citationopen-webui
- HighAny website can steal a signed-in user's session token via the community stats message handleropen-webui
- MediumAny authenticated user can exhaust server memory via a compressed remote image in a chat messageopen-webui
- MediumViewer session token theft via javascript: URLs in chat message linksopen-webui
- MediumInstance-wide stall via automation recurrence rules that evade sub-daily frequency detectionopen-webui
- CriticalMalware in ubiquiti-agents-link-mcpubiquiti-agents-link-mcp
- CriticalMalware in @httttt/mcp-demo@httttt/mcp-demo
- CriticalMalware in @httttt/mcp-npx-fetch-1@httttt/mcp-npx-fetch-1
- HighMCP Atlassian: Arbitrary File Read & Exfiltration (Confused Deputy) in JIRA update_issuemcp-atlassian
- HighMCP Atlassian: ENABLED_TOOLS / Toolset authorization bypassmcp-atlassian
- MediumMCP Atlassian: Insecure File Permissions on OAuth Token Storagemcp-atlassian
- MediumMCP Atlassian: SSRF via DNS Rebinding in Header-Based Authentication Flowmcp-atlassian
- Highmcp-atlassian has an incomplete SSRF remediationmcp-atlassian
- HighMCP Atlassian: SSRF redirect protection missing for basic-auth and OAuth authentication branchesmcp-atlassian
- HighMCP Atlassian: Arbitrary file read via confluence_upload_attachment allows exfiltration of server credentialsmcp-atlassian
- HighMCP Atlassian: Incomplete path traversal fix allows intra-CWD module overwrite and RCE (bypass of GHSA-xjgw-4wvw-rgm4)mcp-atlassian
- Medium@aborruso/ckan-mcp-server has SSRF via DNS-name → internal IP — incomplete fix of CVE-2026-53509@aborruso/ckan-mcp-server
- HighMCP Atlassian: Arbitrary file read/exfiltration via upload_attachment missing validate_safe_path()mcp-atlassian
- HighMCP Atlassian: Unauthenticated arbitrary local file read via upload_attachment file_path, chained with missing auth on streamable-http transmcp-atlassian
- MediumMCP Atlassian: Arbitrary File Read via Upload Attachment Toolsmcp-atlassian
- CriticalMalicious code in ubiquiti-agents-link-mcp (npm)ubiquiti-agents-link-mcp, ubiquiti-agents-link-mcp, ubiquiti-agents-link-mcp, ubiquiti-agents-link-mcp
- HighMCP Atlassian: Arbitrary local file READ via unconstrained file_path in upload_attachment (Confluence + Jira)mcp-atlassian
- HighMCP Atlassian: Arbitrary server-local file upload to Jira/Confluence attachments via unrestricted file_path parametersmcp-atlassian
- MediumMCP Atlassian: Reflected XSS in OAuth Setup Callback Handlermcp-atlassian
- MediumMCP Atlassian: OAuth fallback token storage writes plaintext access and refresh tokens with group-readable permissionsmcp-atlassian
- MediumMCP Atlassian: Path traversal in upload_attachment allows arbitrary file read (incomplete fix for CVE-2026-27825)mcp-atlassian
- HighMCP Atlassian: SSRF Protection Bypassmcp-atlassian
- High@roomi-fields/notebooklm-mcp has a path traversal in vault.batch tool that allows arbitrary file write outside intended vault directory@roomi-fields/notebooklm-mcp
- MediumMCP Atlassian: Path traversal in upload_attachment allows arbitrary file read and exfiltration via MCP tool callmcp-atlassian
- HighMCP Atlassian: HTTP upload tools accept arbitrary server-local file pathsmcp-atlassian
- HighMCP Atlassian: Path Traversal / Arbitrary File Read in confluence_upload_attachment MCP tool (incomplete fix of GHSA-xjgw-4wvw-rgm4)mcp-atlassian
- MediumMCP Atlassian: Incomplete fix for GHSA-7r34-79r5-rcc9: redirect-based SSRF via unhooked requests session in Jira user-permission lookupmcp-atlassian
- HighMCP Atlassian: Jira and Confluence attachment upload tools can read arbitrary server-local filesmcp-atlassian
- HighMCP Atlassian: JIRA_PROJECTS_FILTER / CONFLUENCE_SPACES_FILTER allow forbidden-project content exfiltration (one LIVE-proven on Atlassian Clmcp-atlassian
- Critical[mcp-atlassian] Authentication bypass in HTTP transport: AtlassianOpaqueTokenVerifier accepts any non-empty tokenmcp-atlassian
- HighMCP Atlassian: MCP HTTP Client Server-Local File Exfiltration via Unvalidated Attachment Upload Pathmcp-atlassian
- CriticalMalware in @vite-mcp/vite-type@vite-mcp/vite-type
- CriticalLMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.pylmdeploy
- HighLMDeploy has an SSRF bypasslmdeploy
- HighLMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loadinglmdeploy
- HighObot: Server-Side Request Forgery via remote MCP server URLgithub.com/obot-platform/obot
- MediumObot: MCP Registry API readable without authenticationgithub.com/obot-platform/obot
- HighObot: OAuth Dynamic Client Registration Enables API Token Theft via Audience Confusiongithub.com/obot-platform/obot
- CriticalMalicious code in marketing-mcp (PyPI)marketing-mcp
- MediumvLLM: Request-selected PyNvVideoCodec GPU decode bypasses static VRAM reservationvllm
- MediumRMCP: Custom HTTP headers leak to cross-origin redirect targetsrmcp
- MediumLiteLLM Proxy has server-side request forgery via the `user_config` request parameterlitellm
- HighRMCP: Missing Resource Field Validation in OAuth Protected Resource Metadata Discoveryrmcp
- MediumApproval Bypass in the Send and Wait Node's Approve Within Chat Moden8n, n8n
- HighSQL Injection in Oracle Database Node Delete Table Drop Operationn8n, n8n, n8n
- HighPath Traversal in Signed Resume URL Generation Enables Cross-Project Approval Forgeryn8n, n8n, n8n
- CriticalLMdeploy has Remote Code Execution by Pickle Deserialization via zmq_rpc.call_and_response() in InterLM/lmdeploylmdeploy
- MediumUnescaped Parameter Interpolation into Third-Party Query Languages Enables Filter Bypass and Bulk Data Disclosuren8n, n8n
- HighDuplicate Node IDs Bypass Workflow Credential Tamper Guard, Exposing Credentials to a Shared Workflow Editorn8n
- HighCredential Test Endpoint Resolves Project-Scoped Variables from Attacker-Controlled Project IDn8n, n8n, n8n
- HighPath Traversal in the n8n Node Redirects Public API Calls to Unintended Resourcesn8n, n8n, n8n
- HighRMCP: Unauthenticated permanent session-table leak in rmcp Streamable HTTP server transport leads to remote denial-of-servicermcp
- HighInline Agent Node-Tool Introspection Decrypts Any Instance Credential Without Ownership Checkn8n, n8n
- MediumStored DOM XSS via Resource Locator Dropdown Link Handlingn8n, n8n, n8n
- Critical@zereight/mcp-gitlab: Unauthenticated arbitrary file read via `upload_markdown` enables PAT exfiltration and full account takeover@zereight/mcp-gitlab
- HighCommunity Package Install Validation Bypass via PubSub in Queue Mode Deploymentsn8n, n8n, n8n
- HighWekan and Baserow Credentials Leak Account Password to Unvalidated Host via preAuthentication Hookn8n, n8n
- MediumvLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completionsvllm
- MediumMissing Webhook Signature Verification in Webflow Trigger Node Allows Forged Event Injectionn8n, n8n, n8n
- HighPath Traversal and Query Injection via the Supabase Node Table Namen8n, n8n, n8n
- HighDynamic Credentials Authorize Endpoint Leaks Session Token to Attacker-Controlled Resolvern8n, n8n
- HighNoSQL Injection in MongoDB Chat Memory Node Allows Unauthenticated Cross-Session Chat History Disclosuren8n, n8n, n8n
- HighSupabase Node Filters (String) Mode Allows PostgREST Filter Injectionn8n, n8n, n8n
- Critical@zereight/mcp-gitlab Vulnerable to Server-Side Request Forgery@zereight/mcp-gitlab
- High@zereight/mcp-gitlab has multiple safety-control bypasses: execute_graphql read-only + allow-list bypass, unauthenticated transports, sessio@zereight/mcp-gitlab
- Critical@zereight/mcp-gitlab: DNS rebinding reaches local Streamable HTTP MCP transport@zereight/mcp-gitlab
- MediumDisaggregated generate skips decoder prompt-length validation for some multimodal processorsvllm
- MediumIncomplete artifact pin propagation in FunAudioChat and Tarsier2vllm
- LowSampler Subclass Counter Shadowing Bypasses PyNvVideoCodec Decoder Limits and GPU Memory Accountingvllm
- MediumUncontrolled resource consumption: multimodal chat audio decoding ignores `VLLM_MAX_AUDIO_CLIP_FILESIZE_MB`vllm
- MediumRemote media is fetched and fully materialized before the documented media size / item limits are enforced — remote pre-inference memory andvllm
- MediumRust HTTP/gRPC stop_token_ids bypass Python vocab-bound fix and can terminate EngineCorevllm
- MediumOut-of-range `stop_token_ids` with `min_tokens` can kill vLLM EngineCorevllm
- MediumvLLM DoS via unbounded `cache_salt` length: multi-hundred-MB salt stalls the single EngineCore scheduler thread (CPU exhaustion)vllm
- HighGoogle Meet node commands could skip exec approvalsopenclaw
- MediumMemory dreaming changes could omit owner checksopenclaw
- MediumACP file URLs could skip out-of-cwd read approvalopenclaw
- MediumDiscord guild reads could skip target allowlists@openclaw/discord
- MediumGoogle Meet attendance CSV could interpret names as formulasopenclaw
- HighExec approvals could outlive their reviewed working directoryopenclaw
- MediumBrowser wait predicates could reach blocked destinationsopenclaw
- MediumDiagnostics export could omit owner authorizationopenclaw
- HighGateway upgrade-like requests could retain unauthenticated socketsopenclaw
- Highmessage.action could trust spoofed requester provenanceopenclaw
- HighSlack group DMs could skip sender allowlists@openclaw/slack
- MediumSession filename generation could regain denied toolsopenclaw
- MediumTalk Voice changes could omit admin authorizationopenclaw
- LowMCP loopback could omit sandbox tool deny policyopenclaw
- HighWorkspace Cloud SDK arguments could execute code during Gmail setupopenclaw
- MediumUnicode fallback could escape workspaceOnly rootsopenclaw
- MediumDiscord realtime transcripts could inherit another speaker's owner statusopenclaw
- HighDreaming could widen restricted sender authorityopenclaw
- HighFrontMCP and mcp-from-openapi have bypass of OpenAPI external $ref SSRF fixmcp-from-openapi
- MediumEmbedding fallback could forward a provider key across vendorsopenclaw
- LowTruncated tar listings could bypass extraction limitsopenclaw
- MediumTrusted-host DNS checks could admit unspecified addressesopenclaw
- HighReusable exec approvals could authorize changed argumentsopenclaw
- HighSkill tool dispatch could skip owner-only policyopenclaw
- MediumFile-transfer approvals could widen durable authorityopenclaw
- MediumMicrosoft Teams access-group failures could admit unlisted sendersopenclaw
- MediumTrajectory export could omit owner authorizationopenclaw
- HighCron tool could accept mixed-case command payloadsopenclaw
- HighEscaped newlines could confuse exec allowlist parsingopenclaw
- HighWhatsApp login tool could reach non-owner turns@openclaw/whatsapp
- HighCodex bind could omit owner authorizationopenclaw
- HighBrowser node targets could bypass sandbox host controlopenclaw
- HighSynology file delivery could lose DNS pinningopenclaw
- LowMemory tools could retain revoked accessopenclaw
- HighOpenShell filesystem mutations could race path validationopenclaw
- MediumLINE group allowlist could inherit DM access@openclaw/line
- MediumStructured attachments could hide unvalidated host pathsopenclaw
- MediumVideo asset downloads could reach private destinationsopenclaw
- MediumSynology pre-auth limits could lock out valid webhooksopenclaw
- LowWebhook TaskFlow cancellation could target unrelated sessionsopenclaw
- MediumRevoked Canvas capability could remain active during disconnectopenclaw
- HighExec wrapper allowlists could trust arbitrary inner commandsopenclaw
- MediumFeishu tools could use disabled default accounts@openclaw/feishu
- MediumFeishu unpin could miss cross-context policyopenclaw
- HighMatrix authorization could conflate case-distinct user IDs@openclaw/matrix
- HighAgent cron tool could reach operator command jobsopenclaw
- Mediumsessions.create could miss admin scope for keyed reconfigurationopenclaw
- HighBrowser proxy could miss its admin scopeopenclaw
- MediumBrowser relay authentication could exhaust shared pending capacityopenclaw
- MediumWhatsApp session resets could bypass command authorizationopenclaw
- HighBrowser CDP connections could discard DNS pinningopenclaw
- HighClaude permission replies could omit owner authorizationopenclaw
- MediumWorkspace endpoint override could expose Azure Speech credentialsopenclaw
- HighCodex computer-use install could omit owner authorizationopenclaw
- MediumGroup activation changes could omit owner authorizationopenclaw
- HighConfig revision hashes could expose password verifiersopenclaw
- Highchat.send could expose owner-only infrastructure toolsopenclaw
- HighSignal reactions could bind to the wrong approvalopenclaw
- MediumWindows allowlist execution could use workspace shadowsopenclaw
- MediumQQBot voice filenames could escape the staging directoryopenclaw
- CriticalMySQL MCP Server: Missing Origin/Host Validation in SSE Transport Enables Unauthenticated SQL Execution (DNS Rebinding / Direct Exposure)mysql-mcp-server
- HighInbound voice calls could inherit owner tool authority@openclaw/voice-call
- MediumPrometheus diagnostics could omit operator.read enforcement@openclaw/diagnostics-prometheus
- MediumSlack file downloads could miss conversation authorization@openclaw/slack
- MediumOpenAI-compatible transport could send provider credentials to the wrong endpointopenclaw
- MediumOutbound attachments could ignore requester read denialsopenclaw
- LowActive Memory recall could ignore requester policyopenclaw
- HighCodex native tools could ignore per-chat policyopenclaw
- HighMCP configuration changes could omit owner authorizationopenclaw
- MediumDiscord asset uploads could skip sender media policy@openclaw/discord
- MediumTwilio pre-auth limits could lock out valid webhooksopenclaw
- MediumActive Memory global toggles could omit owner checksopenclaw
- MediumUsage CSV export could interpret session labels as formulasopenclaw
- HighOpen WebUI: Non-admin users can delete admin-owned external knowledge connections via knowledge base deletionopen-webui
- HighCross-Workspace Credential IDORflowise, flowise-components
- Highn8n: Domain-Restriction Bypass via Unguarded Model-Search Endpoint in OpenAI Chat Model Noden8n, n8n, n8n
- HighOpen WebUI: Any authenticated user can reach the Azure platform channel via server-side web fetchopen-webui
- Mediumn8n: Anonymous Approval-Gate Bypass via Reused resumeToken over the Chat WebSocketn8n, n8n
- HighOpen WebUI: Unauthenticated requests can stall the server via uncached OIDC fetches in back-channel logoutopen-webui
- MediumOpen WebUI: Any authenticated user can inject chats into another user's folder via chat completionsopen-webui
- MediumOpen WebUI: Users denied by the OAuth domain allowlist or role policy can still sign in via token exchangeopen-webui
- HighOpen WebUI: SSRF into internal services via DNS rebinding in the Playwright web loaderopen-webui
- Mediumn8n: GitHub Trigger 422 Reuse Path Skips Webhook Secret Storage, Causing Signature Verification to Fail-Openn8n, n8n, n8n
- Mediumn8n: Cross-User Active Workflow ID and Lifecycle Event Disclosure via Missing userId Filtern8n, n8n, n8n
- Mediumn8n: Prototype Pollution via Workflow Structure Summary Can Lead to Denial of Servicen8n, n8n
- Highn8n: Expression Sandbox Escape via Shared Builtin Tampering and Code-Printer Injection Leads to Code Executionn8n, n8n, n8n
- CriticalAuthenticated RCE via MCP Stdio transport allows any user to execute arbitrary OS commands on the serverlangflow
- MediumOpen WebUI: Any authenticated user can start a non-terminating request via a folder parent cycleopen-webui
- CriticalMalware in @yongot/canary-mcp-test-2@yongot/canary-mcp-test-2
- CriticalCross-IdP account takeover via email-only SSO matchingflowise
- Mediumn8n: Cross-Tenant Project-Member PII Disclosure via Missing Per-Project Scope Check on Role Assignment Endpointsn8n, n8n
- CriticalMalware in @yongot/canary-mcp-isolation@yongot/canary-mcp-isolation
- Mediumn8n: Per-Resource OAuth Consent Bypass via Unbound Refresh Token Resource Substitutionn8n, n8n
- Mediumn8n: Path Injection in Elasticsearch and ElasticSecurity Nodes via Unencoded Identifiersn8n, n8n, n8n
- MediumOpen WebUI: Channel members can overwrite another member's message via the chat completions endpointopen-webui
- CriticalMalicious code in fa-mcp-sdk (npm)fa-mcp-sdk
- Highn8n: Unauthenticated Persistent Storage Exhaustion via OAuth Dynamic Client Registration Endpointn8n, n8n
- Highn8n: Expression Sandbox Escape via Class-Field Sanitizer Rebinding Can Lead to Code Executionn8n, n8n, n8n
- Highn8n: Regular Expression Denial of Service in the Default Blocked-File-Pattern Match via a Git Node Clone Pathn8n, n8n, n8n
- HighOpen WebUI: Same-origin XSS to account takeover via terminal port-preview iframe hardcoding allow-same-originopen-webui
- HighMissing authorization on Flowise upsert-history endpoints allows authenticated users and API keys to read and delete arbitrary document-storflowise
- MediumOpen WebUI: A user's session cookies are sent to tool servers configured for bearer authenticationopen-webui
- MediumOpen WebUI: Inaccessible knowledge bases are exposed through the built-in knowledge tool on most vector backendsopen-webui
- Mediumn8n: Disabled OIDC SSO Endpoints Remain Active and Issue Valid Sessionsn8n, n8n, n8n
- HighMissing authorization on Flowise chat message routes allows low-privileged API keys to read and delete chat historyflowise
- Mediumn8n: Log Streaming Event Destinations Decrypt Generic-Auth Credentials Without Ownership Checkn8n, n8n, n8n
- Mediumn8n: Instance AI Credential Setup Accepts Unvalidated Probe URL from Fetched Contentn8n, n8n
- Mediumn8n: Git Node branch.<name>.remote Config Key Bypasses Sandbox Path Restriction, Enabling Local Git Repository Readn8n, n8n, n8n
- CriticalMalware in @yongot/canary-mcp-test@yongot/canary-mcp-test
- HighBullMQ dashboard `/admin/queues` is auth-only (no RBAC), exposing cross-tenant job payloads to every authenticated userflowise
- MediumOpen WebUI: Any authenticated user can suppress calendar alerts instance-wide via a non-numeric alert valueopen-webui
- CriticalINVITED users auto-promoted via SSO email match (invite-token bypass)flowise
- MediumOpen WebUI: Admin demoted through SSO role sync keeps read and write access to all users' notesopen-webui
- HighOpen WebUI: Sign-in as another user via wildcard characters in the OAuth subject claim on SQLiteopen-webui
- MediumOpen WebUI: Users denied by the OAuth role policy can still sign in via token exchangeopen-webui
- MediumOpen WebUI: Any authenticated user can hang the server via message deletion in a cyclic chat treeopen-webui
- CriticalMalicious code in @yongot/canary-mcp-test (npm)@yongot/canary-mcp-test, @yongot/canary-mcp-test, @yongot/canary-mcp-test, @yongot/canary-mcp-test
- MediumOpen WebUI: Server-side fetches reach blocked and internal hosts via unvalidated HTTP redirect targetsopen-webui
- CriticalMalicious code in @yongot/canary-mcp-isolation (npm)@yongot/canary-mcp-isolation, @yongot/canary-mcp-isolation, @yongot/canary-mcp-isolation
- MediumOpen WebUI: Any authenticated user can hang the server via a cyclic chat message historyopen-webui
- Highfunctype-mcp-server: MCP `set_functype_version` Package Alias RCE via Unsanitized pnpm install + Dynamic Importfunctype-mcp-server
- MediumvLLM: SSRF + arbitrary local file read in MiMoV2OmniMultiModalProcessor `_fetch_image` and audio loader bypass MediaConnector protectionsvllm
- Mediumn8n: Agent Workflow Tool Bypasses Sub-Workflow Caller Policyn8n, n8n
- MediumvLLM: Cross-User Data Leak Vulnerabilityvllm
- CriticalMalware in feishu-docx-mcpfeishu-docx-mcp
- CriticalMalicious code in demo-mercadopago-mcp-server (npm)demo-mercadopago-mcp-server, demo-mercadopago-mcp-server, demo-mercadopago-mcp-server, demo-mercadopago-mcp-server, demo-mercadopago-mcp-server, demo-mercadopago-mcp-server
- MediumvLLM: ReDoS via structured_outputs.regex in the lm-format-enforcer backend (no compile timeout) — missed sibling of GHSA-rwxx-mrjm-wc2mvllm
- CriticalMalicious code in create-mcp-use-app (npm)create-mcp-use-app, create-mcp-use-app
- CriticalMalicious code in code-analyzer-mcp (npm)code-analyzer-mcp
- MediumvLLM: Derender endpoints decode caller-supplied GenerateResponse token IDs without output boundsvllm
- CriticalMalicious code in mcp-consultasdeveiculos-client (npm)mcp-consultasdeveiculos-client, mcp-consultasdeveiculos-client, mcp-consultasdeveiculos-client
- MediumvLLM: Unauthenticated Internal Path and Username Disclosure via Validation Error Messagesvllm
- MediumvLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt partsvllm
- HighDenial of service: a negative token id in `/v1/embeddings` or `/pooling` input kills the vLLM engine via a CUDA device-side assertionvllm
- HighOpenClaw Feishu tools could ignore per-account disablement@openclaw/feishu
- LowCKAN MCP Server: Information disclosure via verbose error reflection@aborruso/ckan-mcp-server
- MediumCKAN MCP Server: Cache-key canonicalization collision enables cache confusion / poisoning@aborruso/ckan-mcp-server
- HighOpenClaw Feishu permission tools could ignore per-account disablement@openclaw/feishu
- CriticalMalicious code in amazon-vendor-analytics-mcp-server (npm)amazon-vendor-analytics-mcp-server
- MediumCKAN MCP Server: MQA server allowlist bypass via unanchored regex (`isValidMqaServer`)@aborruso/ckan-mcp-server
- MediumGit Node File Sandbox Escape via Relative Remote URL Base-Directory Mismatchn8n, n8n, n8n
- CriticalMalicious code in auth0-mcp-server (npm)auth0-mcp-server
- MediumImproper Authorization in Source Control Push Allows Cross-Project Workflow and Credential Deletionn8n, n8n, n8n
- CriticalMalicious code in amazon-seller-analytics-mcp-server (npm)amazon-seller-analytics-mcp-server
- HighMLFLOW_ALLOW_PICKLE_DESERIALIZATION=False safety control bypassed by mlflow.statsmodels flavor — RCE via crafted model artifactmlflow
- HighHugging Face Transformers downloads custom generation code before trust consenttransformers
Aug 2026175 advisories · 26 malicious packages · 1 exploited
- HighCross-tenant/cross-org authorization gaps in Enterprise role, and SSO-config endpointsflowise
- HighSSRF bypass via Cheerio/Playwright/Puppeteer document loaders — not covered by httpSecurity deny listflowise, flowise-components
- Highopenai-realtime endpoints allow cross-workspace tool disclosure and execution via unscoped chatflow IDflowise
- CriticalSSRF + LLM API key exfiltration via user-controlled basepath / baseUrl in 8 chat-model nodesflowise, flowise-components
- HighScript injection in Docker image build workflows allows secret theftflowise
- HighCross-tenant organization admin takeover via unscoped organization/workspace membership APIsflowise
- CriticalFlowise RCE via SQL Database Chain Nodeflowise, flowise-components
- CriticalFlowise Custom MCP npx package execution leads to authenticated server-side command executionflowise, flowise-components
- HighMultiple AI Infrastructure Security Vulnerabilities — Credential Exposure, MCP Bypass, Sandbox Escapeflowise
- HighCross-Session Data Leakage via NoSQL Injection in MongoDBMemory Nodeflowise
- CriticalFlowise MCP Server Config cwd Parameter Validation Bypassflowise, flowise-components
- HighLlavaOnevision2 processor loader executes attacker model code with `trust_remote_code=False` (inert `trust_remote_code` kwarg to `transformevllm
- MediumDenial of Service NanoNemoTronVL Video Audio Extraction Bombvllm
- MediumSpeech-to-text audio decode duration limit bypass via forged header sample ratevllm
- MediumBroken access control in GET /api/v1/organizationuser leaks the organization owner's password hash to any member (privilege escalation/accouflowise
- MediumCross-workspace credential IDOR in node-load-method allows low-privilege users to enumerate victim third-party resourcesflowise
- MediumUnauthenticated Denial of Service via text-to-speech/abort Endpoint Allows Termination of Active Chatflow Predictionsflowise
- CriticalLiteLLM vulnerable to server-side template injection in the /prompts/test endpointlitellm
- HighPraisonAI: [Auth Bypass] PraisonAI async Jobs API (`/api/v1/runs`) has no authentication — unauthenticated job execution, result theft, cancpraisonai
- Highmcp-shell has a Secure Mode Allowlist Bypass via Default `/bin/bash` Executablegithub.com/sonirico/mcp-shell
- Highpraisonaiagents web_crawl vulnerable to SSRF via redirect-followingpraisonaiagents
- CriticalPraisonAI has a Browser Server WebSocket origin validation bypass via unanchored regex (patch bypass of CVE-2026-40289 / GHSA-8x8f-54wf-vv92praisonai
- Medium@arikusi/deepseek-mcp-server: Missing Authentication on Self-Hosted HTTP MCP Endpoint@arikusi/deepseek-mcp-server
- Highmcp-shell has a Secure Mode Allowlist Bypass via Git Shell Aliasgithub.com/sonirico/mcp-shell
- Highpraisonaiagents: AgentServer declares auth_token but never enforces it on any routepraisonaiagents
- HighPraisonAI serve agents --api-key is ignored, allowing unauthenticated remote agent executionpraisonai
- Highpraisonaiagents has a `web_crawl` SSRF protection bypass via unchecked redirect targetspraisonaiagents
- Criticalnextcloud-mcp-server: Unauthenticated `POST /webhooks/nextcloud` allows arbitrary vector data deletion when `WEBHOOK_SECRET` is unset ( defanextcloud-mcp-server
- Mediumpraisonaiagents: ast_grep_rewrite rewrites arbitrary files without the @require_approval gate enforced on every sibling mutation toolpraisonaiagents
- HighPraisonAI: [Path Traversal] agent tools escape the configured workspace via symlinkspraisonai
- Highgenieacs-mcp: DNS rebinding reaches local GenieACS MCP Streamable HTTP transportgithub.com/geiserx/genieacs-mcp
- Highmcp-shell — Security Disabled by Default in Bare-Binary Deploy Path + Shell Interpreter in Secure-Mode Allowlistgithub.com/sonirico/mcp-shell
- High@arikusi/deepseek-mcp-server has an Authorization Bypass Through User-Controlled Key@arikusi/deepseek-mcp-server
- HighPraisonAI: Authentication fail-open in Recipe server allows unauthenticated access when API key or JWT auth is configured without a secretpraisonai
- Highpraisonaiagents vulnerable to arbitrary file write via unsanitized `user_id` in `FileMemory.__init__()` — path traversal to any writable locpraisonaiagents
- MediumPraisonAI vulnerable to Server-Side Request Forgery via DNS rebinding bypass in webhook_url validationpraisonai
- HighChainlist has SSRF via MCP SSE and streamable-http transports that allows unauthenticated internal network accesschainlit
- HighPraisonAI workflow include bypasses tools.py autoload opt-in and executes included recipe codepraisonaiagents, praisonai
- Highbrowse-mcp has an arbitrary file write via unconfined download and state pathsbrowse-mcp
- Criticalqwed-mcp has Unsafe SymPy `parse_expr()` Remote Code Execution via Unsanitized Math Expression Inputqwed-mcp
- MediumJupyterHub has Unauthenticated Denial of Service via Unbounded Username Logging on Failed Loginjupyterhub
- HighPraisonAI: Origin-validation bypass (startswith prefix match) enables unauthenticated cross-site request forgery against the PraisonAI MCP Hpraisonai
- HighPraisonAI: `--api-key` flag on `praisonai serve` is not properly enforcedpraisonai
- HighPraisonAI: [Auth Bypass] `praisonai serve agents --api-key` is silently ignored — agent-invocation routes (`POST /agents`, `POST /agents/{agpraisonai
- HighPraisonAI: Webhook SSRF via DNS fail-open in `JobSubmitRequest.validate_webhook_url()` — bypass of CVE-2026-40114praisonai
- Highpraisonaiagents vulnerable to SSRF in web_crawl tool via redirect-following and DNS rebinding (validate-then-fetch gap)praisonaiagents
- Highmcp-contextforge-gateway has Server-Side Template Injection (SSTI) leading to Remote Code Execution in `PromptService._render_template` via mcp-contextforge-gateway
- CriticalChainlit has command injection via MCP stdio transport that allows unauthenticated remote code executionchainlit
- MediumPraisonAI has an origin validation bypass in MCP HTTP Stream transport that allows browser-mediated unauthenticated tool execution on local praisonai
- MediumPraisonAI MCP HTTP server has unauthenticated unbounded session accumulation (memory exhaustion; session TTL never enforced)praisonai
- Highpraisonaiagents has an SSRF protection bypass in `spider_tools._host_is_blocked()` via DNS-resolved hostnames (`127.0.0.1.nip.io`)praisonaiagents
- CriticalMalware in sidecar-mcpsidecar-mcp
- CriticalMalware in mcp-real-chromemcp-real-chrome
- Criticalmcp-contextforge-gateway has RestrictedPython sandbox bypass via getattr builtin in python_sandbox_servermcp-contextforge-gateway
- CriticalXinference vulnerable to remote code execution via unsafe `eval()` in Llama3 tool-call parsingxinference
- CriticalMalware in mcp-server-notionmcp-server-notion
- CriticalMalware in mcp-server-figmamcp-server-figma
- CriticalMalware in mcp-server-githubmcp-server-github
- CriticalMalicious code in @httttt/mcp-demo (npm)@httttt/mcp-demo
- CriticalMalware in mcp-server-sequential-thinkingmcp-server-sequential-thinking
- CriticalMalware in mcp-server-sentrymcp-server-sentry
- CriticalMalware in mcp-server-supabasemcp-server-supabase
- CriticalMalware in mcp-server-gitmcp-server-git
- CriticalMalware in mcp-server-fetchmcp-server-fetch
- CriticalMalware in mcp-server-postgresmcp-server-postgres
- CriticalMalware in mcp-server-redismcp-server-redis
- Mediumlanggraph-api: Relative webhook targets in LangGraph Server can reach in-process routes without authenticationlanggraph-api
- HighContentful MCP Server: export_space/import_space tools pass LLM-controlled `host`/`proxy` args to CMA client, redirecting server PAT to atta@contentful/mcp-server, @contentful/mcp-tools
- HighShared-Workflow Editor Can Exfiltrate Credentials via Workflow Tool Node Inline Sub-Workflown8n, n8n
- CriticalMalicious code in homekit-mcp (npm)homekit-mcp
- CriticalMalicious code in code-assist-mcp (npm)code-assist-mcp
- HighMCP PHP SDK: client HttpTransport SSE buffer (sseBuffer .= chunk) grows unbounded when server withholds the event delimitermcp/sdk
- HighGmail and Brevo nodes accept non-string content, enabling local file read and SSRFn8n, n8n, n8n
- HighExpression Sandbox Escape via $fromAI Prototype Leak Leads to Host RCEn8n, n8n, n8n
- Highgrok-faf-mcp has an arbitrary local file read via unconfined `path` argument in FAF toolsgrok-faf-mcp
- HighExpression Sandbox SpreadElement Bypass Enables Persistent Cross-Evaluation Native Object Mutationn8n, n8n
- MediumSearXNG Basic Authentication Credentials Exposed Through MCP Logs and JSON-RPC Error Responsesmcp-searxng
- Highfaf-mcp has an arbitrary local file read/write via unconfined `path` argument in FAF toolsfaf-mcp
- Mediumlanggraph-api: Incomplete assistant authorization in LangGraph Server run creationlanggraph-api
- MediumInsights API Missing Per-Project Authorization Exposes Workflow Names and Execution Stats Across Projectsn8n, n8n
- MediumLegacy Request Helper SSRF Check Validates uri While Axios Dispatches urln8n, n8n
- HighGit Node Remote Code Execution via Incomplete Repository-Local Configuration Neutralizationn8n, n8n, n8n
- MediumSearXNG MCP Server is Vulnerable to SSRF in web_url_read: the internal-address guard is disabled by default (MCP_HTTP_HARDEN off)mcp-searxng
- CriticalMalicious code in broadcast-graphics-mcp (npm)broadcast-graphics-mcp
- CriticalMalicious code in mcp-dev-toolkit (npm)mcp-dev-toolkit
- Highclaude-faf-mcp has an arbitrary local file read/write via unconfined `path` argument in FAF toolsclaude-faf-mcp
- HighStrapi, SeaTable, and Mailcheck Nodes Leak Decrypted Credential Secrets into Persisted Execution Error Datan8n, n8n, n8n
- CriticalMalicious code in chrome-enterprise-premium-mcp (npm)chrome-enterprise-premium-mcp
- MediumSearXNG MCP Server: Additional hardened-mode SSRF bypassesmcp-searxng
- MediumQuery Injection in Elasticsearch and Google Cloud Firestore Nodes via Unescaped Expression Interpolationn8n, n8n
- CriticalMLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebindingmlflow
- Mediumchrome-devtools-mcp: validatePath() does not canonicalize symlinks before enforcing rootschrome-devtools-mcp
- HighToken Optimizer MCP: OS command injection in smart_user via username in get-user-info@ooples/token-optimizer-mcp
- CriticalMalicious code in @zapier/mcp-integration (npm)@zapier/mcp-integration, @zapier/mcp-integration, @zapier/mcp-integration
- MediumToken Optimizer MCP: Unauthenticated Path Traversal in Dashboard Session Log API Endpoints@ooples/token-optimizer-mcp
- MediumContextForge: DNS TOCTOU race condition causes SSRF protection bypass (`/admin/gateways/test`)mcp-contextforge-gateway
- MediumvLLM: Completion prompt lists fan out into unbounded engine requestsvllm
- MediumPydantic AI: Unvalidated UploadedFile references in UI adapters allow server-side file access using the application's credentialspydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- CriticalMalicious code in @kolbo/mcp (npm)@kolbo/mcp
- CriticalMalicious code in xrblocks-mcp (npm)xrblocks-mcp
- HighMCP-for-Stata: Stata Command Injection via Unsanitized `package` in `ado_package_install`stata-mcp
- CriticalMalicious code in mcp-util-helpers (npm)mcp-util-helpers
- MediumvLLM Unauthenticated Requests Exploit DeepStream Backend Confusion for DoSvllm
- MediumKeras model loading is vulnerable to denial of service through HDF5 shape bombskeras
- CriticalMalicious code in devplatform-react-mcp (npm)devplatform-react-mcp
- MediumLangGraph: Namespace prefix matching crosses segment boundaries in Postgres and SQLite storeslanggraph-checkpoint-postgres, langgraph-checkpoint-sqlite
- CriticalMalicious code in agenthub-multiagent-mcp (npm)agenthub-multiagent-mcp
- MediumEdit Image Node Injection Enables Blind SSRFn8n, n8n, n8n
- HighRCE in the n8n Main Process via Path Traversal in MCP Node-Schema Loadingn8n, n8n
- HighMongoDB Node NoSQL Injection in Find, Delete, and Aggregate Operations via Unescaped Expression Interpolationn8n, n8n, n8n
- HighGraphQL Node Raw Error Re-throw Leaks Decrypted Credential Headers into Persisted Execution Datan8n, n8n, n8n
- MediumSSRF Protection Bypass via SearXNG Tool Allows Member Users to Read Internal Service Responsesn8n, n8n
- LowSSRF Protection Bypass via OAuth2 Credential Token Exchange Reflects Internal Response Bodyn8n, n8n, n8n
- MediumSandbox Escape in JavaScript Code Node via Prototype Pollutionn8n, n8n, n8n
- HighSupabase Node PostgREST Filter Injection in Row Get Many, Delete, and Update Operationsn8n, n8n, n8n
- HighResource Locator Link Preview Expression Injection Allows Cross-User Script Executionn8n, n8n, n8n
- HighMLflow AI Gateway permits SSRF through an unvalidated api_basemlflow
- HighJavaScript Task Runner VM Sandbox Escape via EventEmitter Prototype Pollution Leads to Remote Code Executionn8n, n8n, n8n
- HighCode execution in the n8n Git node via unchecked repository-local git configurationn8n, n8n, n8n
- MediumReDoS in Filter and Switch Node Regex Matching Allows Worker Denial of Servicen8n, n8n, n8n
- HighSnowflake Node Arbitrary File Read and Write via Client-Side Commandsn8n, n8n, n8n
- HighForm Node Completion Page Sandbox CSP Bypass Leads to Stored XSSn8n, n8n, n8n
- MediumMCP create_workflow_from_code Accepts Cross-Project Credentials When Auth Type Is an Expressionn8n, n8n
- MediumGraphQL Node Allowed-Domains Bypass Permits Restricted Credential Exfiltrationn8n, n8n, n8n
- HighCustom-role deletion's reassignment path bypasses project-scoped authorizationn8n, n8n
- MediumFlowise: IDOR vulnerability exists at the GET /api/v1/organization/customer-default-source endpointflowise
- MediumOpen WebUI: Any authenticated user can stall a worker via a knowledge-search pattern that backtracks catastrophicallyopen-webui
- CriticalFlowise: RCE via NodeVM Sandbox Escape in executeJavaScriptCode() nodeVMOptions Overrideflowise, flowise-components
- HighOpen WebUI: A folder write-collaborator can permanently delete the owner's chats by deleting a shared subfolderopen-webui
- MediumLogInputs endpoint bypasses per-run UPDATE authorization in MLflow basic-authmlflow
- MediumOpen WebUI: Tool source code disclosed to read-only users via the tool list and get endpointsopen-webui
- LowOpen WebUI: Any authenticated user can cancel another user's chat generation via the chat delete endpointopen-webui
- HighOpen WebUI: Same-origin XSS to account takeover via terminal file-preview iframe hardcoding allow-same-originopen-webui
- CriticalFlowise: RCE via CSVAgent csvFile data URI base64 segment is interpolated into Python source without validationflowise, flowise-components
- CriticalFlowise: Pyodide validator Unicode homoglyph bypass leads to RCEflowise, flowise-components
- MediumOpen WebUI: Unapproved accounts can open terminal sessions via a WebSocket auth path missing the role checkopen-webui
- CriticalFlowise: CSV Agent Prompt Injection Remote Code Execution Vulnerabilityflowise, flowise-components
- HighFlowise: Unauthenticated Property Injection into Flow Execution Context via Ungated `overrideConfig` Spread in Prediction APIflowise
- MediumOpen WebUI: Cross-user file content disclosure via request-scoped direct model knowledge metadataopen-webui
- MediumOpen WebUI: Instance-wide stall via automation recurrence rules that force multi-second parsingopen-webui
- HighFlowise: Authenticated arbitrary file write in the `S3 Directory` document loader via unsanitized S3 object keysflowise, flowise-components
- MediumFlowise: Unauthenticated Credential Abuse via Text-to-Speech Endpoint Allows Unauthorized Use of Private Chatflow TTS Credentialsflowise
- HighFlowise: RBAC Bypass Leading to Unauthorized Workspace Variables Disclosureflowise
- HighOpen WebUI: Any authenticated user can reach internal services and cloud metadata via NAT64-encoded URLsopen-webui
- CriticalMalware in @servicetitan/anvil2-mcp@servicetitan/anvil2-mcp, @servicetitan/anvil2-mcp, @servicetitan/anvil2-mcp, @servicetitan/anvil2-mcp, @servicetitan/anvil2-mcp, @servicetitan/anvil2-mcp, @servicetitan/anvil2-mcp
- HighFlowise: SSRF Protection Bypass via IPv4-Mapped IPv6 Addressesflowise
- HighFlowise: Cross-workspace credential IDOR in openai-assistants-vector-storeflowise
- HighFlowise: Missing Authorization on Execution Update Endpointflowise
- HighFlowise: Information Disclosure in GET /api/v1/upsert-history returns the entire server-wide upsert historyflowise
- CriticalFlowise RCE via TypeORM DataSourceflowise, flowise-components
- MediumOpen WebUI: Users denied the image-generation permission can still generate images via chat completionsopen-webui
- HighFlowise: Broken Access Control in Stripe Subscription Endpoints Allows Cross-Tenant Billing Manipulationflowise
- MediumCreateModelVersion source validation does not check READ permission on referenced run_idmlflow
- MediumOpen WebUI: DNS Rebinding SSRF Bypassopen-webui
- CriticalMalware in @or-sdk/mcp-tools@or-sdk/mcp-tools, @or-sdk/mcp-tools, @or-sdk/mcp-tools
- MediumOpen WebUI: Deletion of directories and file embeddings in other knowledge bases via sync cleanupopen-webui
- MediumOpen WebUI: Any member with write access to a standard channel can edit or delete other members' messagesopen-webui
- HighFlowise: `DELETE /api/v1/chatflows/:id` does not validate resource type, allowing `agentflows:delete` and `chatflows:delete` to delete each flowise
- HighOpen WebUI: Stored XSS via unescaped KaTeX render-error fallback in rendered messagesopen-webui
- CriticalFlowise: Unauthenticated OAuth2 token refresh endpoint returns access tokens — enables token theft for any connected serviceflowise
- HighFlowise: Unauthenticated OAuth2 Refresh Enables Non-Blind SSRF and Secret Exfiltrationflowise
- MediumOpen WebUI: Client-side SSRF via unrestricted external resource loading in Vega/Vega-Lite chart renderingopen-webui
- HighOpen WebUI: Account takeover via OAuth token exchange accepting tokens issued to any clientopen-webui
- MediumFlowise: Incomplete Credential Redaction Exposes Secrets via APIflowise
- CriticalFlowise: CSV Agent Remote Code Execution via Pyodide Code Injection — Root Shell Verifiedflowise, flowise-components
- HighOpen WebUI: SSRF into internal services via unvalidated sub-resource requests in the Playwright web loaderopen-webui
- HighFlowise: Cross-Workspace OAuth2 Credential Metadata Leakflowise
- CriticalFlowise Sandbox Escape to RCEflowise, flowise-components
- HighFlowise: Missing authorization on `/api/v1/files` allows low-privileged API keys to list and delete files across workspaces within the same flowise
- CriticalFlowise RCE via SQLite Record Manager Nodeflowise, flowise-components
- CriticalFlowise: Remote Code Execution Vulnerability in CSVAgentflowise, flowise-components
- HighFlowise: CVE-2025-8943 Patch Bypass: npm_config_yes bypasses MCP environment variable blocklist (Unauthenticated RCE)flowise, flowise-components
- MediumKeras: HDF5 links can disclose local file contentskeras, keras
- HighTransformers save_pretrained path traversal allows arbitrary file writes through chat template namestransformers
- HighAuthentication Bypass in MCP Composer OAuth Branchlangflow
Jul 2026250 advisories · 72 malicious packages · 1 exploited
- CriticalMalware in gtm-mcp-authgtm-mcp-auth
- CriticalMalware in sap-mcp-configsap-mcp-config
- CriticalMalware in refbase-mcprefbase-mcp
- CriticalMalware in mcp-server-boilerplatemcp-server-boilerplate
- CriticalMalware in pm-claude-skills-mcppm-claude-skills-mcp
- CriticalMalware in iwomm-mcpiwomm-mcp
- CriticalMalware in chaos-mcpchaos-mcp
- CriticalMalware in smart-npv-mcpsmart-npv-mcp
- CriticalMalware in kip-mcp-httpkip-mcp-http
- CriticalMalware in hit-mcphit-mcp
- CriticalMalware in sap-mcp-facilitatorsap-mcp-facilitator
- CriticalMalware in maximumsats-mcpmaximumsats-mcp
- High`@dynatrace-oss/dynatrace-mcp-server` has Unauthenticated HTTP MCP Tool Invocation@dynatrace-oss/dynatrace-mcp-server
- Medium@dynatrace-oss/dynatrace-mcp-server has a DQL injection via parameters not documented as DQL@dynatrace-oss/dynatrace-mcp-server
- CriticalMalware in routerbase-mcprouterbase-mcp
- Medium@dynatrace-oss/dynatrace-mcp-server has a workflow template injection via create_workflow_for_notification@dynatrace-oss/dynatrace-mcp-server
- CriticalMalicious code in mcp-search-server (PyPI)mcp-search-server, mcp-search-server, mcp-search-server
- MediumMCP Ruby SDK: Unbounded session retention in StreamableHTTPTransport allows memory exhaustion via initialize floodmcp
- HighMCP Ruby SDK: Ruby SSE Session Poisoningmcp
- MediumMCP Ruby SDK: Unbounded line buffer in stdio transports leads to memory exhaustion (DoS)mcp
- HighMCP Ruby SDK: Unbounded JSON-RPC request body causes uncontrolled memory allocation in StreamableHTTPTransportmcp
- MediumMCP Ruby SDK: Streamable HTTP transport lacks DNS-rebinding (Host/Origin) protectionmcp
- CriticalCSV_Agent customReadCSV Code Injection Remote Code Execution Vulnerabilityflowise, flowise-components
- CriticalAirtable_Agent Code Injection Remote Code Execution Vulnerabilityflowise, flowise-components
- CriticalFlowise RCE via Custom MCP Config Nodeflowise, flowise-components
- Low@dynatrace-oss/dynatrace-mcp-server's create_dynatrace_notebook missing the human-approval gate@dynatrace-oss/dynatrace-mcp-server
- CriticalFlowise vm2 Sandbox Escape to RCEflowise
- HighAuthenticated Sandbox Escape and Data Exfiltration via Pandas Methods Bypass in pythonCodeValidatorflowise
- HighGitHub MCP Server has Nil Pointer Dereference DoS in completion/complete Handlergithub.com/github/github-mcp-server
- CriticalMalicious code in mcp-server-sentry (npm)mcp-server-sentry, mcp-server-sentry
- CriticalMalicious code in mcp-mermaid (npm)mcp-mermaid, mcp-mermaid
- CriticalMalicious code in proxy-seller-mcp (npm)proxy-seller-mcp
- CriticalMalicious code in mcp-server-redis (npm)mcp-server-redis, mcp-server-redis
- CriticalMalicious code in bitrix24-tasks-mcp-server (npm)bitrix24-tasks-mcp-server, bitrix24-tasks-mcp-server, bitrix24-tasks-mcp-server, bitrix24-tasks-mcp-server, bitrix24-tasks-mcp-server
- CriticalMalicious code in mcp-echarts (npm)mcp-echarts, mcp-echarts
- CriticalMalicious code in @cloudplatform-single-spa/ml-ai-agents-mcp-server (npm)@cloudplatform-single-spa/ml-ai-agents-mcp-server
- CriticalMalicious code in mcp-server-supabase (npm)mcp-server-supabase, mcp-server-supabase
- CriticalMalicious code in brave-search-mcp-server (npm)brave-search-mcp-server
- CriticalMalicious code in mcp-server-postgres (npm)mcp-server-postgres, mcp-server-postgres
- CriticalMalicious code in @djessicatony/folk-mcp-canary (npm)@djessicatony/folk-mcp-canary
- CriticalMalicious code in create-docs-mcp (npm)create-docs-mcp
- CriticalMalicious code in mcp-server-fetch (npm)mcp-server-fetch, mcp-server-fetch
- CriticalMalicious code in sensorium-mcp (npm)sensorium-mcp
- CriticalMalicious code in executable-stories-mcp (npm)executable-stories-mcp
- CriticalMalicious code in @vite-mcp/vite-type (npm)@vite-mcp/vite-type
- CriticalMalicious code in mcp-server-figma (npm)mcp-server-figma, mcp-server-figma
- CriticalMalicious code in mcp-ci-runner-poc (npm)mcp-ci-runner-poc
- CriticalMalicious code in t-invest-mcp-server (npm)t-invest-mcp-server
- CriticalMalicious code in mcp-server-git (npm)mcp-server-git, mcp-server-git
- CriticalMalicious code in polymarket-mcp-v2 (npm)polymarket-mcp-v2
- CriticalMalicious code in zomato-mcp (npm)zomato-mcp
- CriticalMalicious code in mcp-server-github (npm)mcp-server-github, mcp-server-github
- CriticalMalicious code in mcp-server-notion (npm)mcp-server-notion, mcp-server-notion
- CriticalMalicious code in cue-mcp (npm)cue-mcp
- CriticalMalicious code in @antv/mcp-server-antv (npm)@antv/mcp-server-antv, @antv/mcp-server-antv
- CriticalMalicious code in bricks-builder-mcp (npm)bricks-builder-mcp, bricks-builder-mcp, bricks-builder-mcp, bricks-builder-mcp
- CriticalMalicious code in screenpipe-mcp-http (npm)screenpipe-mcp-http
- CriticalMalicious code in mcp-server-sequential-thinking (npm)mcp-server-sequential-thinking, mcp-server-sequential-thinking
- CriticalMalicious code in sidecar-mcp (npm)sidecar-mcp, sidecar-mcp, sidecar-mcp, sidecar-mcp
- HighAWS API MCP Server Security Policy Bypass via Startup Initialization Failureawslabs-aws-api-mcp-server
- MediumOpen WebUI: Model meta.knowledge read-only file access can be upgraded to file write/deleteopen-webui
- LowOpen WebUI: Cross-user model-list exposure via static cache key in get_all_models (aiocache key= vs key_builder= misuse)open-webui
- MediumBlenderMCP path traversal in download_polyhaven_asset allows arbitrary file writesblender-mcp
- HighOpen WebUI: Stored web worker XSS via Pyodideopen-webui
- LowOpen WebUI: Private channel messages can be disclosed through cross-channel thread parent_id bindingopen-webui
- HighOpen WebUI: Cross-user code-interpreter and tool execution via unvalidated Socket.IO event-caller session_idopen-webui
- HighClaude Code: Sandbox Escape via Git Worktree Path Confusion Allows Unsandboxed Code Execution@anthropic-ai/claude-code
- MediumOpen WebUI: Upload `metadata.knowledge_id` bypasses the knowledge-base write-access check (read-only users can add files to KB)open-webui
- MediumOpen WebUI: Account enumeration via observable login timing discrepancyopen-webui
- HighOpen WebUI: Realtime endpoints accept Redis-revoked JWTs after signout/backchannel logoutopen-webui
- MediumOpen WebUI: ReDoS in skill-mention regexes causes whole-instance DoS on default configopen-webui
- Highopen-webui terminal proxy path traversal guard bypass via 9x encoded traversalopen-webui
- MediumOpen WebUI: /api/v1/channels/{id}/members exposes full user model including sensitive credentialsopen-webui
- LowOpen WebUI: Unauthenticated WebSocket Access to Collaborative Document Handlers (ydoc:awareness:update, ydoc:document:leave)open-webui
- HighRay: Arbitrary code execution via ray.data.read_webdataset default decoder: pickle.loads(value) and torch.load(weights_only=False)ray
- HighOpen WebUI: Terminal proxy forwards a spoofable, integrity-unbound user identity to the upstream (X-User-Id header and ws_terminal session_iopen-webui
- HighAWS Bedrock AgentCore: Improper neutralization of argument delimiters in the Python SDK install_packages()bedrock-agentcore
- MediumOpen WebUI: Arena task endpoints can bypass underlying model access controlsopen-webui
- LowOpen WebUI: Scheduled automations continue after pending-user deactivation and stored model ACL revocationopen-webui
- LowONNX: Heap-Buffer-Overflow READ in Gemm Version Converter Adapter via Undersized Input Shapeonnx
- MediumOpen WebUI: `WEB_FETCH_FILTER_LIST` host allow/block filter bypassable via URL path and non-label-boundary matchingopen-webui
- MediumOpen WebUI: POST /api/v1/images/edit bypasses the global image-edit switch and the per-user image-generation permissionopen-webui
- HighOpen WebUI: Cross-channel message overwrite via chat completion API (single-model and multimodel message_ids)open-webui
- CriticalMalware in mcp-notes-server-poc-praetorianmcp-notes-server-poc-praetorian
- Mediumn8n: External Secrets Accessible via Workflow Expressions Outside Credentialsn8n, n8n
- Highn8n: Send Email Node Arbitrary File Read and SSRF via Nodemailer Content-Object Type Confusionn8n, n8n, n8n
- Mediumn8n: Improper Authorization Allows Authenticated Users to Assign Workflows to Folders in Other Projectsn8n
- Mediumn8n: Unauthenticated Endpoint Allows Cancellation of Any User's Active Test Webhookn8n
- Highn8n: SSO Instance-Role Provisioning Allows Privilege Escalation to Instance Ownern8n, n8n, n8n
- LowLiteLLM: Local file read via request-supplied OIDC file referenceslitellm
- MediumLiteLLM: Arbitrary file write via path traversal in Skills archive extractionlitellm
- Highn8n: Bypass "Allowed HTTP Request Domains" Credential Restriction in Multiple AI and LLM Nodesn8n, n8n
- Mediumn8n: Snowflake Node executeQuery Operation Allows SQL Injection via Unparameterized Expression Interpolationn8n, n8n, n8n
- Highn8n: Credential Authorization Bypass via Expression in HTTP Request Node `genericAuthType`n8n, n8n, n8n
- LowLiteLLM: Custom Code Guardrails production endpoints bypass code safety checkslitellm
- HighLiteLLM: MCP Authentication Bypass via OAuth2 Passthrough Fallbacklitellm
- Highn8n: Prototype Pollution via Workflow Credentials Leads to Unauthenticated User and Project Enumerationn8n, n8n, n8n
- Highn8n: Privilege Escalation and Code Execution via Full Public API Key Scope Assignment to Token Exchange JWTsn8n, n8n
- Highn8n: Account Takeover via Unverified Email Claim in Token Exchange Embed Loginn8n, n8n
- Mediumn8n: Custom Header Credential Values Leaked in Plaintext into LLM Node Execution Datan8n, n8n, n8n
- MediumJupyterLab: PyPI extension blocklist package-name canonicalization bypassjupyterlab, jupyterlab
- Mediumn8n: Cross-Tenant Module-Cache Poisoning in the JS Task Runnern8n, n8n, n8n
- Mediumn8n: Google Service Account Private Key Exposed in JWT Headern8n, n8n, n8n
- Mediumn8n: Authenticated SSRF via Dynamic Node Parameters Endpoints Allows Internal Network Accessn8n, n8n, n8n
- Highn8n: Stored DOM XSS via Resource Locator `cachedResultUrl`n8n, n8n, n8n
- Highn8n: Shared-Workflow Editor Can Exfiltrate Credentials via Inline Sub-Workflow JSONn8n, n8n, n8n
- Mediumn8n: computer-use Shell Sandbox Not Enforced on Linux and Windowsn8n, n8n
- Highn8n: Race Condition in Git Clone Node Allows Authenticated Users to Achieve Remote Code Executionn8n, n8n, n8n
- Highn8n: Git Node fetch/pull/pushTags Operations Bypass Sandbox Path Restrictionn8n, n8n, n8n
- Mediumn8n: GraphQL Node Bypasses "Allowed HTTP Request Domains" Restrictionn8n, n8n, n8n
- Highn8n: Expression sandbox escape via arrow-function bodies enabling command executionn8n, n8n
- HighJupyterLab: Image viewer allows XSS when opening malicious image in new browser tabjupyterlab, jupyterlab
- Highn8n: "Allowed HTTP Request Domains" Restriction Bypass via AI Agents MCP Connectorn8n, n8n
- MediumJupyterLab PluginManager lock-rule enforcement bypassjupyterlab, jupyterlab
- Mediumn8n: MySQL v1 Node executeQuery Operation Allows SQL Injection via Unparameterized Expression Interpolationn8n, n8n, n8n
- Mediumn8n: Prototype Pollution via VM Expression Engine Sandbox Escape Leads to Denial of Servicen8n, n8n, n8n
- Mediumn8n: External Secrets Permission Bypass via Expression Parser Mismatchn8n, n8n, n8n
- Mediumn8n: PostgresTrigger Node SQL Injection Allows Authenticated Users to Execute Arbitrary SQL on Connected PostgreSQL Instancesn8n, n8n, n8n
- Highn8n: Cross-Issuer Token Exchange Account Binding via Subject-Only Identity Resolutionn8n, n8n
- Highn8n: DOM-Based XSS via Unsandboxed iframe srcdoc in HTML Previewn8n, n8n, n8n
- Mediumn8n: Path-Confinement Bypass in computer-use search_files Allows Reading Files Outside the Base Directoryn8n, n8n
- Highn8n: Legacy Expression Evaluator Sanitizer Bypass Leads to Authenticated Code Executionn8n, n8n, n8n
- HighJupyterLab: Cross-site scripting (XSS) via crafted settings file (`overrides.json`)jupyterlab, jupyterlab
- Highn8n: Shared Credential Header Leak via HTTP Request Pagination Expressionn8n, n8n, n8n
- Mediumn8n: Member-Level Users Can Execute Other Users' MCP Server Trigger Workflows via Missing OAuth Authorization Checkn8n, n8n
- Highn8n: Authenticated code execution in the n8n Git noden8n, n8n, n8n
- Mediumn8n: SSRF Protection Bypass via MCP Client Noden8n, n8n
- Mediumn8n: Authenticated Users Can Exhaust Temporary Disk Storage via Data-Table File Uploadsn8n, n8n
- LowJupyterLab: Allowlist/blocklist check in `PyPIExtensionManager.install()` not enforced for direct callers (missing `await`)jupyterlab, jupyterlab
- Highn8n: AI Agents Project Viewer Privilege Escalation via run_node_tooln8n, n8n
- Highn8n: Edit Image Node Format Injection Allows Arbitrary File Writen8n, n8n, n8n
- Highn8n: Prototype Pollution via Dot-Notation Field Names Leads To Instance-Wide Denial of Servicen8n, n8n, n8n
- CriticalMalicious code in mcp-runcommand-server (PyPI)mcp-runcommand-server, mcp-runcommand-server, mcp-runcommand-server, mcp-runcommand-server, mcp-runcommand-server, mcp-runcommand-server, mcp-runcommand-server, mcp-runcommand-server
- CriticalMalicious code in mcp-weather-full (PyPI)mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full, mcp-weather-full
- CriticalMalicious code in groq-mcp (PyPI)groq-mcp, groq-mcp
- CriticalMalicious code in mcp-runcommand-server2 (PyPI)mcp-runcommand-server2
- CriticalMalicious code in latinum-wallet-mcp (PyPI)latinum-wallet-mcp, latinum-wallet-mcp, latinum-wallet-mcp, latinum-wallet-mcp, latinum-wallet-mcp
- CriticalMalicious code in litellm (PyPI)litellm, litellm, litellm
- CriticalMalicious code in docontrol-mcp (PyPI)docontrol-mcp
- CriticalMalicious code in xinference (PyPI)xinference, xinference, xinference
- CriticalMalicious code in langchain-core-mcp (PyPI)langchain-core-mcp, langchain-core-mcp
- CriticalMalicious code in mcp-pdftool-plus (PyPI)mcp-pdftool-plus, mcp-pdftool-plus, mcp-pdftool-plus
- CriticalMalicious code in mcp-runcmd-server (PyPI)mcp-runcmd-server, mcp-runcmd-server
- CriticalMalicious code in mistralai (PyPI)mistralai
- CriticalMalicious code in lightning (PyPI)lightning, lightning
- CriticalMalicious code in openai-mcp (PyPI)openai-mcp, openai-mcp
- CriticalMalicious code in ant-mcp-proxy-for-test (PyPI)ant-mcp-proxy-for-test
- CriticalMalicious code in mcp-xyz (PyPI)mcp-xyz
- CriticalMalicious code in story-mcp-hub (PyPI)story-mcp-hub
- CriticalMalicious code in mcp-transport-proto (PyPI)mcp-transport-proto
- CriticalMalicious code in guardrails-ai (PyPI)guardrails-ai
- CriticalMalicious code in vulndify-mcp-server (PyPI)vulndify-mcp-server
- CriticalMalicious code in instructor-mcp (PyPI)instructor-mcp, instructor-mcp
- CriticalMalicious code in tiktoken-mcp (PyPI)tiktoken-mcp, tiktoken-mcp
- CriticalMalicious code in ray-mcp-server (PyPI)ray-mcp-server
- HighvLLM denial of service via prompt embeds on M-RoPE modelsvllm
- HighKeras: TorchModuleWrapper can deserialize unsafe PyTorch pickle datakeras, keras
- Highmeta-ads-mcp: X-Pipeboard-Token Header Auth Bypass Reuses Operator Meta Tokenmeta-ads-mcp
- Highmeta-ads-mcp: Server-Side Request Forgery (SSRF) in `upload_ad_image` via Unrestricted `image_url` Fetchmeta-ads-mcp
- MediumvLLM: Processing differential in multi-channel audio downmixing enables hidden-input/moderation bypass for audio modelsvllm
- HighvLLM has Remote DoS via Invalid Recovered Token Reinjectionvllm
- Mediummcp-memory-keeper: Arbitrary local file read in context_import via unvalidated filePathmcp-memory-keeper
- HighvLLM: ReDoS via structured_outputs.regex compiled without timeout in xgrammar and outlines backendsvllm
- MediumvLLM: Speech-to-text upload size limit is enforced after full UploadFile readvllm
- CriticalGPT Researcher MCP STDIO configuration allows remote command executiongpt-researcher
- HighMCP Python SDK: Experimental task handlers allow any client to access and cancel other clients' tasksmcp
- HighMCP Python SDK: HTTP transports serve session requests without verifying the authenticated principalmcp
- HighMCP Python SDK: WebSocket server transport does not support Host/Origin validationmcp
- CriticalMalware in claude-token-tracker-mcpclaude-token-tracker-mcp
- High@andrea9293/mcp-documentation-server: Web UI API binds to all interfaces without authentication by default@andrea9293/mcp-documentation-server
- HighPyTorch Lightning allows arbitrary code execution through checkpoint _instantiator hyperparameterslightning
- Mediumn8n-MCP: Incorrect authorization can expose default-scope workflow version backups in multi-tenant HTTP moden8n-mcp
- LowKeras: tar extraction permits symlink-based path traversalkeras, keras
- Criticaln8n-MCP: Cross-tenant access to workflow version backups in multi-tenant HTTP deploymentsn8n-mcp
- HighSandbox escape via tampered Python virtual environmentscursor
- HighSandbox escape via launching privileged containerscursor
- HighNetLicensing-MCP: Unauthenticated Use of Server-Side NetLicensing API Key in HTTP Modenetlicensing-mcp
- MediumMCP Atlassian: DNS-rebinding TOCTOU bypass of the SSRF fix (CVE-2026-27826)mcp-atlassian
- Highmcp-atlassian: Arbitrary file read via missing path validation in confluence_upload_attachmentmcp-atlassian
- CriticalMalware in @redhat-cloud-services/hcc-pf-mcp@redhat-cloud-services/hcc-pf-mcp, @redhat-cloud-services/hcc-pf-mcp, @redhat-cloud-services/hcc-pf-mcp
- Highmcp-atlassian: Arbitrary server-side file read via attachment uploadmcp-atlassian
- Criticalmcp-server-kubernetes argument injection can expose Kubernetes cluster credentialsmcp-server-kubernetes
- CriticalMalware in @redhat-cloud-services/hcc-kessel-mcp@redhat-cloud-services/hcc-kessel-mcp, @redhat-cloud-services/hcc-kessel-mcp, @redhat-cloud-services/hcc-kessel-mcp
- MediumGittensory: Missing contributor-scoped access control on profile endpoint and MCP tool leaks miner financial data@jsonbored/gittensory-mcp
- CriticalLocalAI POST /models/apply permits unauthenticated server-side request forgery through gallery URLsgithub.com/mudler/LocalAI
- MediumOpen WebUI allows limited stored XSS vila uploaded html fileopen-webui
- HighOpen WebUI vulnerable to stored XSS via unescaped markdown token in MarkdownTokens.svelte leading to full account takeover and RCE via functopen-webui
- CriticalMalware in mcp-server-pgmcp-server-pg
- Medium@aborruso/ckan-mcp-server: SSRF via base_url allows access to internal networks (Potential fix bypass of CVE-2026-33060)@aborruso/ckan-mcp-server
- MediumONNX has Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)onnx
- MediumOpen WebUI has Blind Server Side Request Forgery in its Image Edit Functionalityopen-webui
- Mediumha-mcp: Add-on settings and policy routes are reachable without authentication at the bare root pathha-mcp
- HighOpen WebUI vulnerable to Stored XSS via iFrame embeds in response messagesopen-webui
- HighOpen WebUI vulnerable to Stored XSS via iFrame in citations modelopen-webui
- CriticalLangroid: Neo4jChatAgent executes LLM-generated Cypher without validation (prompt-to-Cypher injection; config-conditional RCE), mirroring thlangroid
- CriticalLangroid: SQLChatAgent dangerous-function blocklist can be bypassed with quoted or schema-qualified pg_read_file callslangroid
- HighLangroid: handle_message() executes user-supplied tool JSON without sender verificationlangroid
- CriticalLangroid: Sandbox Escape to Remote Code Execution via Incomplete `eval()` Mitigation in TableChatAgentlangroid
- HighKeras: Lambda deserialization can bypass safe mode and execute codekeras, keras
- MediumOpenClaw: Slack and Zalo webhook secrets could remain active after secrets.reloadopenclaw
- HighMLflow: trace API endpoints lack proper authorization validatorsmlflow
- MediumOpenClaw's browser act interactions could bypass private-network navigation checksopenclaw
- HighOpenClaw: Shell wrapper argv could change between approval and executionopenclaw
- HighOpenClaw: Paired nodes could forge exec lifecycle events without system.run provenanceopenclaw
- LowOpenClaw: Feishu dynamic-agent bindings could miss configWrites enforcementopenclaw
- MediumOpenClaw: Mattermost slash token revocation could lag until monitor refreshopenclaw
- MediumOpenClaw: Sandboxed session spawn could expose the real workspace path to child promptsopenclaw
- HighOpenClaw: Hook-triggered CLI runs could receive owner MCP tool authorityopenclaw
- HighOpenClaw: QQBot pre-dispatch slash commands could skip allowFrom checksopenclaw
- HighOpenClaw: Trusted retry endpoint checks could match hostname prefixesopenclaw
- HighOpenClaw: Matrix allowFrom could bind to mutable display namesopenclaw
- MediumOpenClaw: Node pairing reconnection could confuse approval scope stateopenclaw
- Criticalmcp-memory-service: Missing Authentication on Document API Endpoints Allows Unauthenticated Memory Read/Write/Deletemcp-memory-service
- HighOpenClaw: Workspace .env could override Homebrew executable selection for skill install flowsopenclaw
- HighOpenClaw: Slack allowFrom could bind to mutable display namesopenclaw
- HighOpenClaw: Control UI locality spoofing could mint a durable admin device tokenopenclaw
- MediumOpenClaw: Skill Workshop apply flow could override pending approvalopenclaw
- HighGrackle: Fail-open authorization in the MCP tool layer lets scoped agents perform cross-task and cross-session mutations (IDOR)@grackle-ai/mcp
- HighLangroid: Path traversal in the file tools allows read/write outside configured current directorylangroid
- MediumOpenClaw: Mattermost handlers could fall open when channel type was missingopenclaw
- MediumOpenClaw: message.action forwarding could send Gateway credentials to model-supplied loopback URLsopenclaw
- MediumOpenClaw: Browser debug/export routes could reuse already-open blocked tabsopenclaw
- HighOpenClaw: Scoped chat.send route inheritance could bypass admin command scope gatesopenclaw
- HighOpenClaw: PowerShell encoded-command aliases could miss exec allowlist checksopenclaw
- HighOpenClaw: QQBot streaming command could mutate config without explicit allowFromopenclaw
- HighOpenClaw: QQBot native approval buttons did not enforce configured approver identityopenclaw
- HighOpenClaw's POSIX node system.run safe-bin allowlist could be widened by shell expansionopenclaw
- MediumOpenClaw: memory-wiki ingest could read local files with operator.write scopeopenclaw
- MediumOpenClaw: Embedded runner policy could be confused by provider aliasesopenclaw
- HighOpenClaw: Native command authorization could skip owner-command enforcementopenclaw
- HighLangroid: SQLChatAgent _validate_query blocklist misses pg_read_file family enabling arbitrary file readlangroid
- HighOpenClaw: Message read actions could skip channel allowlist checksopenclaw
- MediumOpenClaw: Bundle MCP loopback could miss its exec denylist on session spawnopenclaw
- HighOpenClaw: Trusted-proxy Control UI WebSocket accepted client-declared scopes before pairingopenclaw
- HighOpenClaw: Same-host trusted-proxy deployments could accept local forged identity headersopenclaw
- MediumOpenClaw: MCP loopback could skip owner-only tool policy for non-owner callersopenclaw
- Criticalfast-mcp-telegram: Bearer token path traversal bypasses reserved Telegram session protectionfast-mcp-telegram
- HighOpenClaw's marketplace runtime extension metadata could point at unscanned payloadsopenclaw
- HighOpenClaw: Fake package roots could influence memory-core artifact loadingopenclaw
- HighOpenClaw: Combined POSIX shell options could confuse exec revalidationopenclaw
- CriticalOpenClaw: QQBot admin commands could skip DM-only and allowFrom policyopenclaw
- HighOpenClaw: Telegram interactive callbacks could skip commands.allowFromopenclaw
- MediumOpenClaw's Slack plugin approvals used the exec approver gate for plugin actionsopenclaw
- HighOpenClaw: Non-owner chat senders could issue device-pairing bootstrap codesopenclaw
- HighOpenClaw: Exec approval display truncation could hide the command being approvedopenclaw
- MediumKeras: HDF5 virtual datasets can disclose local fileskeras, keras
- HighApify Model Context Protocol (MCP) server: Actor MCP path authority injection leaks Apify token@apify/actors-mcp-server
- MediumOpenClaw MCP SSE redirects could forward Authorization headersopenclaw
- HighGradio FileExplorer preprocess path traversal allows files outside root_dir to reach callbacksgradio
- Highauth-fetch-mcp has SSRF Protection Bypass via IPv4-mapped IPv6 Loopbackauth-fetch-mcp
Jun 2026234 advisories · 8 malicious packages · 1 exploited
- HighBrowser snapshot routes could miss post-navigation SSRF checksopenclaw
- HighExec allowlist glob matching could allow traversal bypassesopenclaw
- Highflock wrapper could bypass durable exec approval bindingopenclaw
- HighDiscord guild actions could skip cross-provider requester authorizationopenclaw
- CriticalMalware in autotel-mcpautotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp, autotel-mcp
- HighWorkspace dotenv files could override provider credentialsopenclaw
- MediumRemote media URLs could slow-read exhaust tool workersopenclaw
- HighMCP loopback could expose owner-only tools to non-owner runsopenclaw
- HighSandbox bind mounts could bypass parent-directory denylist checksopenclaw
- MediumWebSocket auth attempts could avoid non-browser rate limitsopenclaw
- HighHooks allowedAgentIds could be bypassed with blank agent IDsopenclaw
- CriticalMalware in autotel-mcp-instrumentationautotel-mcp-instrumentation, autotel-mcp-instrumentation, autotel-mcp-instrumentation, autotel-mcp-instrumentation, autotel-mcp-instrumentation, autotel-mcp-instrumentation
- HighQQBot exec approvals could allow non-allowlisted sendersopenclaw
- HighPlugin install commands could allow non-owner persistenceopenclaw
- Medium[AgentGG] MS Teams allowFrom could bind to mutable display names@openclaw/msteams
- HighNode exec approvals could use different gateway and node environmentsopenclaw
- Highdevice.pair.approve could bypass role-management checksopenclaw
- HighHost exec environment filtering could allow Git ext transportopenclaw
- HighIdentity-bearing HTTP callers could reach admin-scoped toolsopenclaw
- HighDiscord moderation actions could miss trusted requester checksopenclaw
- HighWhatsApp group IDs could satisfy elevated sender allowlistsopenclaw
- MediumClickClack allowFrom could allow non-allowlisted commandsopenclaw
- MediumQQBot media upload could reach untrusted remote URLsopenclaw
- HighHost exec environment filtering could miss interpreter startup variablesopenclaw
- HighDevice-pair approval could expose node system.run earlyopenclaw
- MediumTrajectory export could skip broad credential redactionopenclaw
- HighOpenAI-compatible HTTP model overrides could miss admin authorizationopenclaw
- HighOpenShell mirror sync could follow remote symlink parentsopenclaw
- HighSandbox exec-server HTTP requests could reach internal networksopenclaw
- MediumSkill command dispatch could skip effective tool policyopenclaw
- HighIsolated cron jobs could regain denied exec toolsopenclaw
- HighMS Teams message actions could miss requester authorizationopenclaw
- MediumBot Framework serviceUrl validation could leak bot tokens@openclaw/msteams
- HighSetup-mode discovery could load untrusted workspace pluginsopenclaw
- MediumNative web search could ignore OpenClaw tool policyopenclaw
- MediumMS Teams outbound requests could leak Bot Framework tokens@openclaw/msteams
- CriticalLangflow: Code Injection Vulnerability in Code Validation Endpointlangflow
- HighMessage mutations could skip requester authorizationopenclaw
- HighHTTP Canvas responses could forge trusted A2UI actionsopenclaw
- MediumPlugin install wrappers could skip install policyopenclaw
- HighClickClack agent-mode dispatch could ignore toolsAllowopenclaw
- HighHost exec environment filtering could miss rustup startup variablesopenclaw
- HighMS Teams safeFetch could race DNS rebinding checksopenclaw
- HighBrowser act route could miss current-tab URL checksopenclaw
- CriticalMCP Toolbox HTTP tool path parameters allow traversal to unintended endpointsgithub.com/googleapis/mcp-toolbox
- Highmcp-memory-service: OAuth read-only clients can write and delete memories through MCP tools/callmcp-memory-service
- HighStreamable HTTP mode exposes LINE Desktop read/send tools without MCP authenticationline-desktop-mcp
- Criticalmcp-pinot: Unauthenticated tool invocation via default oauth_enabled=False + host 0.0.0.0 bindmcp-pinot-server
- Mediumjupyter/nbconvert allows Cross-site Scripting (XSS) via unsanitized `text/vnd.mermaid` output in HTML exportsnbconvert
- Mediumpydantic-ai: SSRF blocklist bypass via IPv4-compatible, SIIT/IVI, and local NAT64 IPv6 addresses (incomplete fix of CVE-2026-46678)pydantic-ai, pydantic-ai, pydantic-ai-slim, pydantic-ai-slim
- Low@cardano402/mcp-server missing spending limits, LAN-exposed HTTP transport, and SSRF via catalog.server.url@cardano402/mcp-server
- Medium@anthropic-ai/claude-code has an Insecure Temporary File in /copy Command that Enables Response Disclosure and Symlink-Based File Write@anthropic-ai/claude-code
- MediumLangGraph Checkpoint: Unsafe JSON deserialization in checkpoint loadinglanggraph-checkpoint
- MediumGitHub MCP Server: Lockdown mode singleton in HTTP server causes cross-user GraphQL client confusiongithub.com/github/github-mcp-server
- MediumLangGraph SDK has unsafe URL path constructionlanggraph-sdk
- MediumJenkins MCP Server Plugin missing a permission checkio.jenkins.plugins:mcp-server
- Mediumforeman-mcp-server Inserts Sensitive Information into Log Fileforeman-mcp-server
- CriticalChainlit contains a session hijacking vulnerabilitychainlit
- MediumKeras: DiskIOStore permits path traversal through crafted layer nameskeras, keras
- HighApache Doris MCP Server is vulnerable to SQL Injection via metadata query pathdoris-mcp-server
- MediumLiteLLM: MCP Proxy Has Improper Authenticationlitellm
- LowLiteLLM: Admin Key Handler Has Improper Authorizationlitellm
- LowBerriAI litellm has SSRF via Unvalidated spec_path URL in MCP OpenAPI Spec Loaderlitellm
- MediumLiteLLM: SSO Debug Flow Has Improper Authenticationlitellm
- LowBerriAI litellm: UI User Enumeration leads to System-Wide Information Disclosurelitellm
- LowLiteLLM: PROXY_ADMIN database API Key Generator Has Insufficient Session Expirationlitellm
- LowBerriAI litellm has Security Feature Bypass in BannedKeywords and AzureContentSafety Guardrails via call_type Mismatch on Async Endpointslitellm
- LowLiteLLM: M2M JWT Handler Has Improper Authorizationlitellm
- LowBerriAI litellm: SSO Login Does Not Invalidate Previous UI Session Tokenslitellm
- HighImproper neutralization of argument delimiters in AWS Bedrock AgentCore Python SDK install_packages()bedrock-agentcore
- MediumLangflow: Logout button does not clear sessionlangflow
- CriticalLangflow: BaseFileComponent-based nodes arbitrary file read with RCE exploitlangflow
- HighLangSmith SDK TracingMiddleware: Arbitrary server-side file readlangsmith
- Mediumdbt MCP Server: Unauthenticated OAuth Context Endpoint Leaks dbt Platform Tokensdbt-mcp
- HighSearXNG MCP Server: DNS-resolved Private Hostname SSRF in `web_url_read`mcp-searxng
- HighLangflow: IDOR Vulnerability in `/api/v1/responses` Endpoint Allows Authenticated Attackers to Access Another User's Flowlangflow
- HighLangflow: Unauthenticated DoS through multipart form boundary file uploadlangflow
- HighKozou: Unauthenticated MCP HTTP server and bundled dev-stack hardening (DNS-rebinding, request-body limits, read-only reads, default network@kozou/mcp
- MediumJupyterLab: Stored XSS in extension manager through package metadata unsanitized URI protocoljupyterlab
- Highappium-mcp: Unescaped Locator Data XSS in MCP-UI Resource (createLocatorGeneratorUI)appium-mcp
- HighSearXNG MCP Server: Unbounded Response Body Read Bypasses URL Size Limit in `web_url_read`mcp-searxng
- HighPraisonAI Dynamic Context history and terminal tools read files outside configured storage via path traversalpraisonai
- HighOpenClaw: Workspace .env npm_execpath could influence bundled runtime dependency installopenclaw
- CriticalPraisonAI: Arbitrary File Read/Write via `multiedit` Tool Without Path Validationpraisonai
- HighPraisonAI: Arbitrary File Read via `@file:` Mention Path Traversalpraisonaiagents
- MediumPraisonAI: Unauthenticated Event Injection via SSE `/publish` Endpointpraisonaiagents
- CriticalPraisonAI: Unauthenticated RCE via Jobs API + Approval Bypasspraisonaiagents, praisonai
- Criticalgemini-mcp-tool vulnerable to OS command injection and @file exfiltration via prompt quoting (CVE-2026-0755)gemini-mcp-tool
- MediumOpenClaw: Internal/webchat command auth could inherit ownerAllowFrom wildcard stateopenclaw
- HighPraisonAI: Server-Side Request Forgery (SSRF) in SearxNG / search_web tools via attacker-controlled searxng_url parameterpraisonaiagents
- HighOpenClaw: Shell positional parameters could weaken strict inline-eval checksopenclaw
- HighMCP Toolbox for Databases: authenticated authorization bypassgithub.com/googleapis/mcp-toolbox
- HighPraisonAI recipe serve Typer command bypasses the non-localhost authentication guardpraisonai
- HighPraisonAI: HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous toolspraisonai
- LowOpenClaw: Skill-command dispatch could skip before-tool-call hooksopenclaw
- MediumPraisonAI: SpiderTools redirect-target SSRF protection bypasspraisonaiagents
- HighPraisonAI SandlockSandbox falls back to unrestricted subprocess execution when Landlock is unavailablepraisonai
- MediumOpenClaw: memory-wiki shared search could miss session visibility checksopenclaw
- HighPraisonAI recipe workflow policy can be bypassed by declaring and YAML-approving dangerous tools outside TEMPLATE.yamlpraisonai
- HighPraisonAI DiscordApproval accepts unrelated channel messages as dangerous-tool approvalspraisonai
- CriticalPraisonAI: AgentOS remains unauthenticated after incomplete fix version and allows remote agent invocationpraisonai
- HighOpenClaw: Zalo allowFrom could bind to mutable display namesopenclaw
- HighPraisonAI: PRAISONAI_CALL_AUTH=disabled environment variable unconditionally disables authenticationpraisonai
- Criticalgoogleapis/mcp-toolbox: authentication bypass vulnerability in the generic opaque token validation path (validateOpaqueToken)github.com/googleapis/mcp-toolbox
- LowOpenClaw: BlueBubbles sender policy could match mutable conversation identifiersopenclaw
- Mediummarimo contains a reflected cross-site scripting vulnerability in the notebook pagemarimo
- LowOpenClaw: Empty-scope device re-pairing could confuse caller scope containmentopenclaw
- MediumOpenClaw: Tool group policy callers could accept unvalidated group IDsopenclaw
- LowOpenClaw: Bootstrap token replay could widen pending pairing scopesopenclaw
- MediumOpenClaw: macOS Swift exec allowlist missed combined POSIX inline flagsopenclaw
- HighPraisonAI: IMAP Command Injection via Unsanitized Email Search Parameterspraisonaiagents
- HighOpenClaw: Host environment sanitizer missed two Node.js control variablesopenclaw
- HighOpenClaw: Discord allowFrom could bind to mutable display namesopenclaw
- LowOpenClaw: Exec allowlist could miss side effects from transparent command wrappersopenclaw
- HighOpenClaw: Shell inline-command parsing could miss an allowlist checkopenclaw
- HighPraisonAI GitHub template cache path traversal allows outside-cache file write and directory deletionpraisonai
- HighPraisonAI LinearBot processes unsigned webhooks when LINEAR_WEBHOOK_SECRET is missingpraisonai
- MediumOpenClaw: Slack reaction events could ignore reaction notification settingsopenclaw
- CriticalJupyter Server: Stored XSS in `NbconvertFileHandler` / `NbconvertPostHandler` via missing `sandbox` CSPjupyter-server
- CriticalPraisonAI: Jobs API exposes agent-execution endpoints with no authenticationpraisonai
- HighOpenClaw: Workspace .env CLOUDSDK_PYTHON could influence Gmail setup gcloud executionopenclaw
- HighPraisonAI Code agent tools fail open without a workspace boundarypraisonai
- MediumOpenClaw: Hostname checks could treat trailing-dot hosts inconsistentlyopenclaw
- Criticalnetlicensing-mcp: REST Path Traversal Bypasses Token Redactionnetlicensing-mcp
- Criticalpraisonai: recipe serve auth middleware silently disables itself when no secret is setpraisonai
- HighPraisonAI dynamic-context artifact tools read arbitrary host files outside artifact storagepraisonai
- HighPraisonAI A2U incomplete authentication fix leaves current serve command unauthenticated by defaultpraisonai
- MediumOpenClaw: Focus command could miss controlScope enforcementopenclaw
- HighPraisonAI: Unauthenticated Local File Inclusion via agent_file path in PraisonAI Jobs APIpraisonai
- CriticalPraisonAI: Missing Authentication for Critical Function and Improper Neutralization of Special Elements used in an OS Command ('OS Command Ipraisonai
- MediumPraisonAI: execute_code sandbox bypass: str.format C-level attribute access reads every blocklisted dunderpraisonaiagents
- HighOpenClaw: Pairing-scoped device session could restore revoked node token authorityopenclaw
- HighPraisonAI Slack app_mention bypasses configured user/channel authorizationpraisonai
- HighDocker MCP Gateway: Argument injection via OCI image label YAMLgithub.com/docker/mcp-gateway
- HighPraisonAI: Jobs webhook SSRF protection bypass via DNS rebindingpraisonai
- MediumOpenClaw: Config recovery could restore openclaw.json with broad file permissionsopenclaw
- HighOpenClaw: Workspace-derived service PATH could influence trash command selectionopenclaw
- HighOpenClaw: Linux and macOS exec allowlists skipped configured argument patternsopenclaw
- HighPraisonAI recipe.run_stream skips dangerous-tool policy enforcementpraisonai
- HighPraisonAI ToolsMCPServer legacy SSE transport accepts attacker Host/Origin and exposes registered toolspraisonaiagents, praisonai
- Highpraisonaiagents: SSRF guard validates literal IPs only and never resolves DNSpraisonaiagents
- HighPraisonAI: Compute-bridged file tools allow shell command injectionpraisonai
- MediumOpenClaw: Exported session HTML could keep unsafe markdown linksopenclaw
- HighOpenClaw: Workspace .env STATE_DIRECTORY could influence bundled runtime dependency rootsopenclaw
- Criticalgoogleapis/mcp-toolbox: authentication bypass vulnerability in the generic opaque token validation path (validateOpaqueToken)github.com/googleapis/mcp-toolbox
- CriticalPraisonAI: MCP SSE transport binds 0.0.0.0 with no authentication and no Origin validation; bundled SecurityConfig is never wired inpraisonaiagents
- MediumOpenClaw: Active Memory write scope could mutate global configopenclaw
- CriticalPraisonAI AgentTeam.launch exposes unauthenticated remote agent listing and invocation endpointspraisonaiagents
- HighPraisonAI: Webhook signature verification skipped (fail-open) when secret unset, allowing forged inbound webhooks (WhatsApp & Linear bots)praisonai
- HighOpen WebUI: Redirect-Bypass SSRF in OAuth `_process_picture_url` (incomplete-fix sibling of CVE-2026-45401)open-webui
- MediumChrome DevTools for agents: daemon.pid write follows symlinks in /tmp fallback runtime directorychrome-devtools-mcp
- HighOpen WebUI: Cross-origin postMessage confirmation bypass via action:submitopen-webui
- MediumOpen WebUI Prompt history IDOR: unbound history_id allows cross-prompt read and deletionopen-webui
- MediumvLLM: GGUF dequantize kernel int truncation exposes uninitialized GPU memory in multi-tenant servingvllm
- CriticalMalware in @mastra/mcp-registry-registry@mastra/mcp-registry-registry
- Mediumn8n: Wrong OAuth Scope on Evaluation Test Runs Endpointsn8n, n8n, n8n
- MediumvLLM: OOM Denial of Service via Audio Decompression Bombvllm
- MediumvLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernelsvllm
- MediumOpen WebUI: Any authenticated user can read other users' private notes via Socket.IOopen-webui
- MediumvLLM: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectationsvllm
- MediumOpen WebUI: Authenticated users can target arbitrary configured Ollama backends via unguarded url_idx path parameteropen-webui
- MediumOpen WebUI BOLA: `search_knowledge_files` Allows Unauthorized Knowledge Base File Enumerationopen-webui
- MediumOpen WebUI IDOR: Calendar event re-parenting allows writing events into another user's calendaropen-webui
- MediumClaude Code: Out-of-Band Data Exfiltration via Pre-Approved HuggingFace Domain in WebFetch@anthropic-ai/claude-code
- MediumvLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic routervllm
- MediumOpen WebUI: Sibling-Prefix Path Traversal via /cache/{path}open-webui
- HighOpen WebUI: SSRF Protection Bypass in Playwright Web Loader via HTTP Redirectsopen-webui
- MediumOpen WebUI: RAG ACL Bypass in Milvus Multitenancy Modeopen-webui
- CriticalMalware in @mastra/core@mastra/core
- HighOpen WebUI: Path traversal / SSRF in terminal server proxy via encoded path traversalopen-webui
- HighOpenClaw: MCP Streamable HTTP redirects could forward configured custom headers to another originopenclaw
- HighOpen WebUI: Stored XSS to Account Takeover via Model Profile Imagesopen-webui
- HighOpen WebUI: Stored XSS in Mermaid Markdown Previewopen-webui
- HighOpen WebUI: Forged model meta.knowledge allows cross-user file read and deletionopen-webui
- HighOpen WebUI: Forged chat-file link allows cross-user file read and deletionopen-webui
- MediumOpen WebUI: Cross-user file disclosure via /api/chat/completions image_url fieldopen-webui
- CriticalMalware in @mastra/mcp@mastra/mcp
- CriticalMalware in @mastra/mcp-docs-server@mastra/mcp-docs-server
- CriticalLangflow: Unauthenticated file upload leads to DoS (space exhaustion) and information leaklangflow
- Highn8n: Cross-Tenant Credential Takeover via Dynamic Credentials EE Endpointsn8n, n8n, n8n
- Mediumn8n: Prototype Pollution enables confused-deputy execution via public webhooksn8n, n8n
- Highn8n: Stored XSS in Chat Trigger Noden8n, n8n, n8n
- CriticalLiteLLM: Authentication Bypass via Host Header Injectionlitellm
- Mediumn8n: Git Node Clone and Push Operations Bypass File Sandboxn8n, n8n, n8n
- MediumLangflow: Path Traversal in Knowledge Bases API via Creation Endpointlangflow
- CriticalvLLM: OpenAI auth bypassvllm
- Mediumn8n: Merge Node SQL Mode Prototype Pollutionn8n, n8n
- HighLangflow: IDOR/BOLA in Monitor API — Missing Ownership Enforcement on 7 Endpointslangflow
- Highn8n: Python sandbox escapen8n, n8n, n8n
- Mediumn8n: SQL Injection in Postgres v1/TimesclaeDB Nodesn8n, n8n
- MediumLangChain: Path traversal and sandbox escape in LangChain file-search middleware and loaderslangchain
- Mediumn8n: Public API Execution Retry Authorization Bypassn8n, n8n
- Mediumn8n: Reflected XSS via Facebook, WhatsApp, and Microsoft Teams Trigger Webhook Verification Endpointsn8n
- CriticalMalware in sp-api-dev-assistant-mcp-serversp-api-dev-assistant-mcp-server
- Mediumn8n: Wrong OAuth Scope On Evaluations Test Run Creation Endpointn8n, n8n, n8n
- Mediumn8n: NoSQL Injection in MongoDB Node Find And Replace Operationn8n
- Mediumn8n: Denial of Service via ZIP decompression in webhook workflown8n
- Mediumn8n: Missing Token Validation on Microsoft Agent 365 Trigger and Stripe Nodesn8n, n8n
- Mediumn8n: Python Code Node AST Validator Bypassn8n, n8n
- Highn8n: Credential Exfiltration via Permission Bypassn8n, n8n, n8n
- HighvLLM: Security Check Bypass via assert Statement in Activation Function Loading Allows Arbitrary Code Executionvllm
- Highn8n: MCP Browser HTTP Transport Exposes Unauthenticated Browser-Control Sessionsn8n, n8n
- MediumLangflow: Unauthenticated Shareable Playground arbitrary local or S3 file readlangflow
- Highn8n: SecurityScorecard Node Leaks API Token to User-Controlled Hostn8n, n8n, n8n
- CriticalLangflow: Unauthenticated RCE in Shareable Playgroundslangflow
- Highn8n: Same-Origin XSS in Respond to Webhook Noden8n, n8n, n8n
- Highn8n: Microsoft SQL Node Prototype Pollutionn8n
- CriticalMCP Toolbox for Databases has an Origin Validation Errorgithub.com/googleapis/mcp-toolbox
- HighChromaDB allows any authenticated users to arbitrarily read, write, update, or delete data in any tenant's collectionchromadb
- CriticalChromaDB has a code injection vulnerabilitychromadb
- HighChromaDB's SimpleRBACAuthorizationProvider doesn't check which tenant, database, or collection a permission applies tochromadb
- CriticalMeta Ads MCP: Unauthenticated HTTP MCP Tool Execution Leaks Operator Meta Access Tokenmeta-ads-mcp
- HighKeras archive extraction utilities allow path traversal and arbitrary file writeskeras
- HighvLLM is vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.loavllm
- MediumvLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processorsvllm
- LowWeaviate has an Improper Authorization issuegithub.com/weaviate/weaviate
- LowMySQL MCP Server is Vulnerable to SQL Injection Through its mysql URI Handlermysql-mcp-server
- CriticalCursor Desktop sandbox escape via agent-controlled working directorycursor
- CriticalCursor Desktop sandbox escape via symlink and failed path canonicalizationcursor
- MediumMCP Server Kubernetes: kubectl-generic flag injection enables Kubernetes bearer token exfiltrationmcp-server-kubernetes
- CriticalMCP-for-Stata: Command injection via log_file_name parameter in Stata command wrapperstata-mcp
- LowMLflow: Deterministic sampling in dataset digest enables predictable collisionsmlflow
- LowGradio: Audio cache key ignores metadata when saving numpy audio outputsgradio
- Lowmilvus: RBAC grantee-id uses truncated MD5 (64-bit), enabling privilege-binding collisions and cross-role privilege forgerygithub.com/milvus-io/milvus
- LowSGLang is Vulnerable to DoS via the data_hash Functionsglang
- LowStreamlit @st.cache_data hash collision via fixed sampling seed and PIL P-mode palette omissionstreamlit
- LowCode Index MCP is vulnerable to Uncontrolled Resource Consumptioncode-index-mcp
- Highhuggingface/transformers: Arbitrary Code Execution During Model Initialization in the LightGlue Model Loading Pathtransformers
- CriticalMLflow: Environment variable injection in AI Gateway secrets enables server-side credential exfiltrationmlflow
- LowFoundationAgents MetaGPT: Deserialization through flawed argument mapping via Message.check_instruct_content()metagpt
- LowSGLang: Reachable Assertion via lora_path in LoRAManager enables remote Denial of Dervicesglang
- MediumMLflow: Any authenticated user can enumerate all gateway secrets, endpoints, and model definitionsmlflow
- High@agenticmail/mcp Missing Authentication for Critical Function@agenticmail/mcp
- CriticalMalware in @redhat-cloud-services/hcc-feo-mcp@redhat-cloud-services/hcc-feo-mcp, @redhat-cloud-services/hcc-feo-mcp
May 2026214 advisories · 7 malicious packages · 1 exploited
- LowAider is vulnerable to Code Injection via editor_coder.run functionaider-chat
- LowAider has an SSRF vulnerability through its AWS EC2 Metadata Endpointaider-chat
- CriticalPraisonAI vulnerable to sandbox escape via `print.__self__` builtins module leak in `execute_code` (subprocess mode)praisonaiagents, praisonai
- MediumPraisonAI spider_tools SSRF protection bypass via alternate loopback host encodingspraisonaiagents, praisonai
- MediumPraisonAI CLI automatically resolves @url mentions in prompt text and can read loopback URLs into model contextpraisonaiagents, praisonai
- HighPraisonAI: Arbitrary code execution via unguarded `spec.loader.exec_module` in `agents_generator.py` - sibling of CVE-2026-44334praisonai
- Highagno contains a SQL injection vulnerabilityagno
- CriticalPraisonAI `deploy --type api` emits a Flask server with authentication disabled by defaultpraisonai
- CriticalPraisonAI call server exposes unauthenticated agent listing, invocation, and deletion when CALL_SERVER_TOKEN is unsetpraisonai
- HighPraisonAI vulnerable to unauthenticated arbitrary file read via MCP workflow.show, workflow.validate, deploy.validatepraisonai
- HighPraisonAI has an Arbitrary File Write in Python APIpraisonai
- CriticalPraisonAI's unauthenticated A2A official example can reach real LLM-driven `eval()` tool executionpraisonai
- CriticalMCP Toolbox for Databases vulnerable to DNS rebinding attacksgithub.com/googleapis/mcp-toolbox
- HighGradio contains a cookie injection vulnerabilitygradio
- CriticalLangroid has Prompt to SQL Injection, Leading to RCElangroid
- HighHuggingFace transformers vulnerable to remote code executiontransformers
- CriticalMLflow allows unauthorized access to multipart upload endpoints when the `--serve-artifacts` mode is enabledmlflow
- Mediumvllm has Improper Resource Shutdown or Releasevllm
- CriticalMalware in explorhub-mcp-serverexplorhub-mcp-server
- CriticalMalware in mcp-dashboards-sharedmcp-dashboards-shared
- Highlmdeploy: Hardcoded trust_remote_code=True is an implicit unsafe remote-code load path with no user opt-outlmdeploy
- MediumPydantic AI: SSRF cloud-metadata blocklist bypass via IPv4-mapped IPv6 (Incomplete fix of CVE-2026-25580)pydantic-ai, pydantic-ai-slim
- HighMCP Server Kubernetes: Tool Access Control Bypass via Presentation-Layer Filtering Without Execution-Layer Enforcementmcp-server-kubernetes
- HighLMDeploy: Arbitrary code execution via hardcoded trust_remote_code=True in lmdeploy model initializationlmdeploy
- HighCursor Desktop sandbox escape via Claude hook configurationcursor
- HighLiteLLM allows an authenticated internal_user to create API keys with access to routes that their role does not permitlitellm
- HighWindows-MCP: HTTP transports expose unauthenticated PowerShell control with wildcard CORSwindows-mcp
- MediumMLflow authenticated users can enumerate any registered model versions due to lack of per-model permissions checksmlflow
- HighLiteLLM allows a user to modify their own user_role via the /user/update endpointlitellm
- MediumFlowise: Mass Assignment in PUT /api/v1/user Allows Authenticated Users to Override Password Hash and Bypass Password Change Verificationflowise
- HighDiffusers: TOCTOU Trust Remote Code Bypassdiffusers
- MediumFlowise: Cross-Workspace Chatflow Disclosure via chatflows/apikey Endpoint Returns All Unprotected Chatflowsflowise
- MediumFlowise: Hardcoded CORS wildcard on TTS endpoint enables cross-origin credential abuse from any webpageflowise
- HighPenPot MCP REPL server binds to 0.0.0.0 with unauthenticated /execute endpoint — RCE@penpot/mcp
- LowMCP Registry: OCI validator skips ownership check on upstream rate limitsgithub.com/modelcontextprotocol/registry
- Mediumn8n: Legacy ExecuteWorkflow Node Bypassed File Path Restrictionsn8n
- Mediumn8n: Credential exfiltration via Allowed HTTP Request Domains Bypassn8n
- CriticalMLflow: Improper Origin Validation in MLflow Assistant /ajax-api Endpoints Enables Browser-Mediated Local Command Executionmlflow
- HighCamoFox MCP: Unauthenticated HTTP MCP browser-control surfacecamofox-mcp
- CriticalMCP Gateway: Authority-injection and JWT/session bypass via the unauthenticated router hair-pin "router-key" / "mcp-init-host" pathgithub.com/Kuadrant/mcp-gateway
- Highauth-fetch-mcp: SSRF and disk exfiltration via unvalidated auth_fetch and download_media URLsauth-fetch-mcp
- MediumApify Model Context Protocol (MCP) server: Domain Allowlist Bypass in fetch-apify-docs via String Prefix Matching@apify/actors-mcp-server
- CriticalMalware in @antv/mcp-server-chart@antv/mcp-server-chart, @antv/mcp-server-chart
- CriticalMalicious code in guardrails-ai 0.10.1 (supply chain compromise)guardrails-ai
- CriticalSGLang: Unauthenticated RCE via --enable-custom-logit-processorsglang
- Low@ai-sdk/provider-utils has an Uncontrolled Resource Consumption issue@ai-sdk/provider-utils, @ai-sdk/provider-utils, @ai-sdk/provider-utils
- HighMLFlow Creates a Temporary File With Insecure Permissionsmlflow
- Mediumn8n-MCP: Workflow telemetry sanitizer could retain partial values from URL-shaped node parametersn8n-mcp
- CriticalChromaDB Python project has a pre-authentication code injection vulnerabilitychromadb
- CriticalSGLanG: Multimodal scheduler deserializes untrusted pickle data on 0.0.0.0 ROUTER socketsglang
- Highn8n-MCP: Multi-tenant MCP requests fall back to process-level n8n credentials when tenant headers are absent or incompleten8n-mcp
- HighSpring AI MCP Security: Unvalidated URL Fetching (SSRF)org.springaicommunity:mcp-client-security
- CriticalSGLang's multimodal generation runtime has an unauthenticated path traversal vulnerabilitysglang
- CriticalMalicious dropper in mistralai 2.4.6 PyPI packagemistralai
- HighMLflow: unauthenticated access to certain FastAPI routesmlflow
- HighOpen WebUI Vulnerable to SSRF via OAuth Profile Picture URL in _process_picture_url (oauth.py)open-webui
- HighOpen WebUI: shared-chat branch ignores access_type, allowing unauthorized file deletionopen-webui
- HighOpen WebUI has XSS via SVG in /api/v1/channels/webhooks/{webhook_id}/profile/imageopen-webui
- HighOpen WebUI vulnerable to stored XSS via OAuth picture claim stored as SVG data URI in profile_image_urlopen-webui
- HighOpen WebUI: Jupyter code execution works despite `ENABLE_CODE_EXECUTION=false` — feature gate bypassedopen-webui
- HighOpen WebUI Vulnerable to IDOR: Retrieval API Bypasses Knowledge Base Access Controlsopen-webui
- HighOpen WebUI's chat completion API allows tool restrictions to be bypassedopen-webui
- HighOpen WebUI has a full SSRF Vulnerability in the RAG Web Search Featureopen-webui
- HighOpen WebUI has stored XSS via the HTML renedering viewopen-webui
- Criticaln8n Has an Arbitrary File Read via Git Noden8n, n8n, n8n
- MediumOpen WebUI's API key endpoint restrictions bypassed via `x-api-key` header — full message processing on restricted endpointsopen-webui
- MediumOpen WebUI has an IDOR vulnerability in the pin_channel_message API endpointopen-webui
- HighFlowiseAI: Dataset create+update mass-assignment allows cross-workspace dataset takeoverflowise
- HighFlowiseAI has Mass Assignment in Chatflow Update Endpoint that Allows Cross-Workspace AgentFlow Reassignmentflowise
- MediumOpen WebUI Vulnerable to Unauthenticated RAG Configuration Disclosureopen-webui
- HighFlowiseAI has Mass Assignment in Variable Update Endpoint that Allows Cross-Workspace Resource Reassignmentflowise
- MediumOpen WebUI has Stored Cross-Site Scripting In Profile Pictureopen-webui
- Highn8n Has a Cross-user Authorization Bypass in Dynamic Credential OAuth Endpointsn8n, n8n, n8n
- HighFlowiseAI: CustomTemplate create+update mass-assignment allows cross-workspace template takeoverflowise
- HighFlowiseAI: Assistant create+update mass-assignment allows cross-workspace assistant takeoverflowise
- HighFlowiseAI Vulnerable to Credential Data Leakflowise
- HighFlowiseAI: DatasetRow create+update mass-assignment allows cross-workspace row takeoverflowise
- Lowdbt MCP Server Logs Tool Arguments Including SQL Queries and Credentials in Plaintext Without Redaction When File Logging Is Enableddbt-mcp
- HighOpen WebUI: Low-privilege authenticated users can enumerate and stop global background tasks, causing system-wide chat disruptionopen-webui
- HighOpen WebUI has a Server-Side Request Forgery (SSRF) bypass in `validate_url`open-webui
- CriticalFlowiseAI: Authenticated Host RCE via POST /api/v1/node-custom-function and NodeVM Sandbox Escapeflowise
- Criticaln8n: HTTP Request Node Pagination Prototype Pollution to RCEn8n, n8n, n8n
- HighOpen WebUI has Stored XSS in Banner Component via Improper Sanitization Orderopen-webui
- MediumOpen WebUI vulnerable to blind server side request forgery (SSRF) via the PDF generate functionopen-webui
- MediumHome Assistant MCP Server: YAML config backups written under www/ are served unauthenticated at /local/ha-mcp
- HighOpen WebUI has Broken Access Control for Completions APIopen-webui
- MediumOpen WebUI missing authorization check at the model update function - models from other users can be updatedopen-webui
- MediumOpen WebUI: Sharing models for others to use (read permission) also exposes model details (system prompt leakage)open-webui
- HighOpen WebUI: LDAP and OAuth First-User Race Condition Allows Multiple Admin Accountsopen-webui
- MediumOpen WebUI has stored XSS via unsanitized Office/Excel/DOCX file preview rendering ({@html} without DOMPurify)open-webui
- HighFlowiseAI: Vector Store No Permission Checksflowise
- HighFlowiseAI has Mass Assignment in Assistant Update Endpoint that Allows Cross-Workspace Resource Reassignmentflowise
- MediumOpen WebUI Vulnerable to Cross-Site Request Forgery (CSRF) via Image URL Manipulationopen-webui
- MediumOpen WebUI Exposes System Prompt to Regular User [Non-Admin]open-webui
- Lowdbt MCP Server Transmits All MCP Tool Arguments Including Raw SQL and --vars Credentials to dbt Labs Telemetry by Default Without Redactiondbt-mcp
- LowOpen WebUI: Read-Only Users Can Toggle Note Pin Status via Incorrect Permission Check (Write via Read-Only Access)open-webui
- MediumOpen WebUI: Unauthenticated endpoint can trigger embedding generation (cost/DoS)open-webui
- HighOpen WebUI has stored XSS via attacker-controlled file extension in /api/v1/audio/transcriptionsopen-webui
- HighFlowise has an MCP Security Bypass that Enables RCEflowise, flowise-components
- Highn8n Has a Source Control Pull SQL Injectionn8n, n8n, n8n
- HighFlowiseAI: Evaluation create+update mass-assignment allows cross-workspace evaluation takeoverflowise
- HighOpen WebUI: Missing `workspace.tools` Authorization Check on Tool Update Endpoint Allows Privilege Escalation to Code Executionopen-webui
- HighFlowiseAI Exposes Basic Auth Credentials via APIflowise
- MediumOpen WebUI Has Stored Cross-Site Scripting in SVG Rendereropen-webui
- HighOpen WebUI: Cross-User File Access via Unchecked file_id in Folder Knowledge and Knowledge-Base Attach Endpointsopen-webui
- HighOpen WebUI: Missing permission check in files API allows authenticated users to list, access and delete every uploaded fileopen-webui
- HighOpen WebUI has a SSRF Bypass via HTTP Redirect Following in Web-Fetch and Image-Load Endpoints (not addressed by CVE-2025-65958)open-webui
- MediumOpen WebUI: Mass Assignment via FeedbackForm extra=allow Allows Feedback User ID Spoofing and Evaluation Data Manipulationopen-webui
- MediumOpen WebUI: Authenticated users can bypass model access control via exposed query parameter [AI-ASSISTED]open-webui
- Criticaln8n Has an XML Node Prototype Pollution Patch Bypassn8n, n8n, n8n
- MediumOpen WebUI has an IDOR vulnerability in the update_message_by_id API endpointopen-webui
- HighFlowiseAI: Evaluator create+update mass-assignment allows cross-workspace evaluator takeoverflowise
- MediumOpen WebUI has an Indirect Object Reference (IDOR) in user notesopen-webui
- HighFlowiseAI has Mass Assignment in Tool Update Endpoint that Allows Cross-Workspace Resource Reassignmentflowise
- Mediumdbt MCP Server has an Argument Injection in dbt CLI Tool Wrappers via node_selection and resource_type Parametersdbt-mcp
- HighLangSmith SDK: Public prompt pull deserializes untrusted manifests without trust boundary warninglangsmith, langchain
- CriticalObot has an authorization bypass in /mcp-connect/{id} that allows any authenticated user to use any registered MCP servergithub.com/obot-platform/obot
- HighSpring AI: Prompt Injection via Memory Poisoning in PromptChatMemoryAdvisororg.springframework.ai:spring-ai-client-chat, org.springframework.ai:spring-ai-client-chat
- HighPyTorch Lightning load_from_checkpoint has an insecure checkpoint deserializationpytorch-lightning
- CriticalOpen Source Kubectl MCP Server vulnerable to arbitrary code execution via user interaction with crafted HTML pagekubectl-mcp-server, kubectl-mcp-server
- Mediummem0 server lacks authentication and authorization controls for its memory creation API endpointmem0ai
- CriticalMalware in nextmove-mcpnextmove-mcp, nextmove-mcp, nextmove-mcp, nextmove-mcp, nextmove-mcp
- Mediummem0 server lacks authentication and authorization controls for its memory deletion API endpointmem0ai
- CriticalMalware in cmux-agent-mcpcmux-agent-mcp, cmux-agent-mcp, cmux-agent-mcp, cmux-agent-mcp, cmux-agent-mcp, cmux-agent-mcp
- Highmem0 server lacks authentication and authorization controls for its memory management API endpointsmem0ai
- HighSpring AI: ChatMemory DEFAULT_CONVERSATION_ID causes unintended cross-user data leakageorg.springframework.ai:spring-ai-client-chat, org.springframework.ai:spring-ai-client-chat, org.springframework.ai:spring-ai-client-chat, org.springframework.ai:spring-ai-model, org.springframework.ai:spring-ai-model, org.springframework.ai:spring-ai-model
- CriticalGuardrails AI contains a code injection vulnerability in its Hub package installation mechanismguardrails-ai
- CriticalMalware in @squawk/mcp@squawk/mcp, @squawk/mcp, @squawk/mcp, @squawk/mcp, @squawk/mcp
- MediumPraisonAI knowledge-store backends interpolate unvalidated collection names into SQL and CQL queriespraisonai
- HighMLflow allows an unauthenticated remote attacker to read arbitrary files from the server's filesystemmlflow
- HighMLflow Has a Server-Side Request Forgery (SSRF) Vulnerabilitymlflow
- HighPraisonAI ships and generates a legacy API server with authentication disabled by default, allowing unauthenticated workflow executionpraisonai
- HighOpen WebUI has a CORS misconfiguration and session validation issueopen-webui
- HighBentoML Dockerfile command injection via docker.base_image (sister of pending GHSA-w2pm-x38x-jp44 / CVE-2026-33744 / CVE-2026-35043)bentoml
- HighGitHub Copilot CLI: Nested Bare Repository Can Execute Arbitrary Commands via core.fsmonitor@github/copilot
- CriticalPraisonAI MCP `tools/call` path-traversal => RCE via Python `.pth` injectionpraisonai
- HighPraisonAI's symlink-extraction bypass of `_safe_extractall` writes outside `dest_dir`praisonai
- HighPraisonAI has unsafe tool resolution in `ToolExecutionMixin.execute_tool`: undeclared `__main__` callables executepraisonaiagents, praisonai
- HighOpen WebUI has inconsistent authorization controls within memories APIopen-webui
- HighOpen WebUI Arbitrary File Write, Delete via Path Traversalopen-webui
- MediumOpen WebUI's Improper Authorization in Standard Channels Allows Message Updates with Read Permissionopen-webui
- HighOpen WebUI's Insecure Message Access Breaks Authorizationopen-webui
- HighDockerfile command injection via envs[*].name in bentofile.yaml (sibling fix-bypass of CVE-2026-33744 and CVE-2026-35043)bentoml
- HighLiteLLM has a sandbox escape in custom-code guardraillitellm
- High@yoda.digital/gitlab-mcp-server's SSE transport has no authentication and wildcard CORS, exposing all 86 GitLab tools@yoda.digital/gitlab-mcp-server
- Critical@profullstack/mcp-server vulnerable to OS Command Injection in domain_lookup Module@profullstack/mcp-server
- CriticalOpen WebUI has an LDAP Empty Password Authentication Bypassopen-webui
- HighOpen WebUI: Redis Cache Keys tool_servers and terminal_servers Missing Instance Prefix Enable Cross-Instance Cache Poisoningopen-webui
- HighOpen WebUI: Stale Admin Role in Socket.IO Session Pool Enables Post-Demotion Cross-User Note Accessopen-webui
- HighOpen WebUI has Improper Authorization Controlopen-webui
- Highgmaps-mcp's unauthenticated HTTP transport allows unlimited Google Maps API calls at operator expensegmaps-mcp
- MediumOpen WebUI vulnerable to Global Knowledge Base Enumeration via knowledge-bases Meta-Collectionopen-webui
- HighOpen WebUI has Knowledge Base Destruction and RAG Poisoning via Unauthorized Collection Overwriteopen-webui
- MediumOpen WebUI's Channel Access Grants Bypass filter_allowed_access_grantsopen-webui
- Highn8n-mcp affected by path traversal, redirect-following SSRF, and telemetry payload exposuren8n-mcp
- LowMCP Registry's GitHub OIDC tokens are replayable across registry deployments due to shared audiencegithub.com/modelcontextprotocol/registry
- HighOpen WebUI Vulnerable to Arbitrary File Upload and Path Traversalopen-webui
- HighOpen WebUI's Base Model Routing Bypasses Access Control via Model Chainingopen-webui
- MediumOpen WebUI Missing Access Check on Channel Members Endpoint for Standard Channelsopen-webui
- Highn8n-mcp webhook and API client paths has an authenticated SSRFn8n-mcp
- MediumOpen WebUI has Stored XSS in Pending User Overlay via Incorrect DOMPurify Application Orderopen-webui
- Highopen-webui Vulnerable to Stored XSS via Model Descriptionopen-webui, open-webui
- MediumOpen WebUI has Unauthorized File and Knowledge Base Content Access via RAG Vector Searchopen-webui
- MediumOpen WebUI: Deactivated Channel Members Retain Full Access to Group/DM Channelsopen-webui
- HighOpen WebUI's responses passthrough endpoint lacks access control authorizationopen-webui
- MediumOpen WebUI's Mass Assignment via Pydantic extra='allow' Allows Creating Folders in Other Users' Accountsopen-webui
- HighOpen WebUI has stored XSS in Excel file previewopen-webui
- MediumOpen WebUI's Model Import Overwrites Any Model Without Ownership Checkopen-webui
- HighLangChain vulnerable to unsafe deserialization of attacker-controlled objects through overly broad `load()` allowlistslangchain-core, langchain-core
- MediumMCP Registry has an unauthenticated SSRF: HTTP namespace verification dials 6to4 / NAT64 / site-local IPv6 addresses, bypassing private-addrgithub.com/modelcontextprotocol/registry
- MediumOpen WebUI's Ollama Model Access Control Bypass via /api/generate, /api/embed, /api/embeddings, and /api/showopen-webui
- MediumMCP Registry vulnerable to stored XSS in catalogue UI via attribute-quote breakout in publisher-controlled `websiteUrl`github.com/modelcontextprotocol/registry
- MediumMCP Registry has open redirect via protocol-relative path in trailing-slash middlewaregithub.com/modelcontextprotocol/registry
- MediumRead-Only Open WebUI Users Can Modify Collaborative Documents via Socket.IOopen-webui
- HighDiffusers has a `trust_remote_code` bypass via `custom_pipeline` and local custom componentsdiffusers
- Highmcp-ssh-tool has file transfer path policy bypass and bearer token comparison hardeningmcp-ssh-tool
- MediumBentoML has Information Disclosure in `bentoml build` via symlink traversal in the build contextbentoml
- CriticalCompromise of PyTorch Lightning PyPi Package Versionspytorch-lightning, pytorch-lightning
- MediumvLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parametersvllm
- Highrmcp Streamable HTTP server transport has a DNS rebinding vulnerabilityrmcp
- MediumFlowise: Bcrypt Password Hash Exposureflowise
- HighKeras vulnerable to DoS via Malicious .keras Model (HDF5 Shape Bomb Causes Petabyte Allocation in KerasFileEditor)keras, keras
- HighJupyterLab's command linker attributes in HTML enable one-click command execution from untrusted contentjupyterlab, notebook
- HighPraisonAI has an SSRF bypasspraisonaiagents
- HighPraisonAI has unauthenticated RCE via `tool_override.py` (CVE-2026-40287 patch bypass)praisonai
- HighJupyter Server has a CORS Origin Validation Bypass via `re.match()` in `allow_origin_pat`jupyter-server
- MediumOpenClaw contains a symlink traversal vulnerabilityopenclaw
- HighJupyterLab has an Extension Manager API/GUI Policy Discrepancy, allowing 3rd party (malicious) extensions install via POST requestjupyterlab
- Mediumwireshark-mcp vulnerable to arbitrary file write via export_objects when WIRESHARK_MCP_ALLOWED_DIRS is not configuredwireshark-mcp
- HighJupyter Server: Path Traversal via incorrect startswith() root directory check allows access to sibling directoriesjupyter-server
- HighJupyter Server's Authentication Cookies Remain Valid After Password Reset and Server Restartjupyter-server
- Mediumogham-mcp had credentials embedded in published PyPI sdists -- Neon postgres URLs and Voyage API keyogham-mcp
- CriticalLangflow Knowledge Bases API is Vulnerable to Path Traversallangflow
- HighOpenClaw's gateway config mutation guard allowed unsafe model-driven config writesopenclaw
- MediumvLLM Vulnerable to Remote DoS via Special-Token Placeholdersvllm
- MediumJupyterHub has cross-origin form POSTs bypass XSRF (CWE-352)jupyterhub
- MediumOpenClaw's Webhooks SecretRef route secret remains valid after rotation/reloadopenclaw
- MediumJupyter Server has an open redirection vulnerability in `next` query parameterjupyter-server
- HighOpenClaw vulnerable to arbitrary code execution via attacker-controlled setup-api.js loaded from cwd during env-key resolutionopenclaw
- MediumOpenClaw validates Zalo outbound photo URLs through the SSRF guardopenclaw
- MediumOpenClaw: Workspace dotenv files cannot override connector endpoint hostsopenclaw
- MediumOpenClaw: OpenShell FS bridge reads pin and verify the opened file before returning bytesopenclaw
- MediumOpenClaw's Gateway Control UI bootstrap config required Gateway authopenclaw
- CriticalMalware in pi-exa-mcppi-exa-mcp
- Low@puchunjie/doc-tools-mcp has a Path Traversal Issue@puchunjie/doc-tools-mcp
- MediumOpenClaw's ACP child sessions inherit subagent security envelope constraintsopenclaw
- LowOpenClaw: Slack thread context could include messages from non-allowlisted sendersopenclaw
- HighOpenClaw: MCP loopback owner context is derived from server-issued bearer tokensopenclaw
- MediumOpenClaw: OpenShell FS bridge writes stay pinned to the sandbox mount rootopenclaw
- MediumOpenClaw's exec allowlist analysis rejects shell expansion in unquoted heredocsopenclaw
- HighOllama contains a heap out-of-bounds read vulnerability in the GGUF model loadergithub.com/ollama/ollama
- MediumSGLang has an Improper Input Validation/Injection Issuesglang
- Lowyii2-mcp-server has a Command Injection Issueyii2-mcp-server
- Lowmem0ai mem0 has an Improper Input Validation Issuemem0ai
Apr 2026340 advisories · 2 malicious packages · 3 exploited
- Highn8n-mcp's IPv4-mapped IPv6 addresses bypass SSRF protection in validateUrlSync(), enabling full SSRF for SDK embeddersn8n-mcp
- Mediummcp-server-semgrep has a Command Injection issuemcp-server-semgrep
- HighJupyter Notebook Vulnerable to Authentication Token Theft via CommandLinker XSSjupyterlab, notebook
- Highn8n has a Python Task Runner Sandbox Escape Vulnerabilityn8n, n8n, n8n
- Highn8n Vulnerable to Unauthenticated Denial of Service via MCP Client Registrationn8n, n8n, n8n
- Highn8n Vulnerable to XSS via MCP OAuth clientn8n, n8n, n8n
- Mediumn8n has Public API Variables IDOR that Allows Cross-Project Secret Disclosuren8n, n8n, n8n
- MediumOpenClaw: Owner-enforced commands could accept wildcard channel senders as command ownersopenclaw
- Mediumn8n has Open Redirect in MCP OAuth Consent Flown8n, n8n, n8n
- Mediumn8n Vulnerable to Hijacking of Unauthenticated Chat Executionn8n, n8n, n8n
- MediumOpenClaw: Webchat audio embedding could read local files without local-root containmentopenclaw
- Mediumn8n has SQL Injection in Snowflake and MySQL Nodesn8n, n8n, n8n
- Criticaln8n has XML Node Prototype Pollution that to RCEn8n, n8n, n8n
- Mediumn8n has SQL Injection in SeaTable Noden8n, n8n, n8n
- MediumClaude SDK for TypeScript has Insecure Default File Permissions in Local Filesystem Memory Tool@anthropic-ai/sdk
- Criticaln8n has Prototype Pollution in XML Webhook Body Parser that Leads to RCEn8n, n8n, n8n
- Highn8n's Credential Authorization Bypass in dynamic-node-parameters Allows Foreign API Key Replayn8n, n8n
- Mediumn8n has SQL Injection in Oracle Database Node via Limit Fieldn8n, n8n, n8n
- Mediumsqlite-mcp has an Injection issuesqlite-mcp
- HighSpring AI has a VectorStore FilterExpression Converter injectionorg.springframework.ai:spring-ai-vector-store, org.springframework.ai:spring-ai-vector-store
- Mediumnotes-mcp has a Path Traversal issuenotes-mcp
- Mediummkdocs-mcp-plugin has a Path Traversal issuemkdocs-mcp-plugin
- Mediummcp-url-downloader has a Server-Side Request Forgery issuemcp-url-downloader
- Mediumkaggle-mcp has a Path Traversal issuekaggle-mcp
- LowvLLM makes Use of Uninitialized Resourcevllm
- LowOllama is Vulnerable to Path Traversalgithub.com/ollama/ollama
- MediumOpenClaw: Hook mapping templates could bypass hook session-key opt-inopenclaw
- LowOpenClaw: Isolated cron awareness events were recorded as trusted system eventsopenclaw
- MediumOpenClaw: Feishu card actions could misclassify DMs and skip dmPolicyopenclaw
- MediumOpenClaw: Agent gateway config mutations could change protected operator settingsopenclaw
- MediumOpenClaw: QQBot direct media upload skipped URL SSRF validationopenclaw
- MediumOpenClaw: Workspace dotenv MiniMax host override could redirect credentialed requestsopenclaw
- HighOpenClaw: Workspace dotenv could override runtime-control environment variablesopenclaw
- LowOpenClaw: Browser CDP profile creation skipped strict-mode SSRF checksopenclaw
- MediumOpenClaw: MCP stdio server env could load dangerous startup variables from workspace configopenclaw
- MediumOpenClaw: Bundled MCP/LSP tools could bypass configured tool policyopenclaw
- HighLiteLLM: Authenticated command execution via MCP stdio test endpointslitellm
- LowOpenClaw: Assistant media route missed scope enforcement for trusted-proxy authorizationopenclaw
- Mediumn8n-MCP: Sensitive MCP tool-call arguments logged on authenticated requests in HTTP moden8n-mcp
- LowOpenClaw: Paired-device pairing actions were not limited to the caller deviceopenclaw
- HighRay: Remote Code Execution via Parquet Arrow Extension Type Deserializationray
- HighClaude Code: Trust Dialog Bypass via Git Worktree Spoofing Allows Arbitrary Code Execution@anthropic-ai/claude-code
- CriticalLiteLLM has SQL Injection in Proxy API key verificationlitellm
- CriticalGemini CLI: Remote Code Execution via workspace trust and tool allowlisting bypasses@google/gemini-cli, @google/gemini-cli
- HighLiteLLM: Server-Side Template Injection in /prompts/test endpointlitellm
- Mediumn8n-MCP Logs Sensitive Request Data on Unauthorized /mcp Requestsn8n-mcp
- CriticalMalware in claudcode-mcpclaudcode-mcp
- CriticalFlowise: CSV Agent Prompt Injection Remote Code Execution Vulnerabilityflowise, flowise-components
- Mediumnbconvert has an Arbitrary File Write via Path Traversal in Cell Attachment Filenamesnbconvert
- HighLMDeploy has Server-Side Request Forgery (SSRF) via Vision-Language Image Loadinglmdeploy
- Mediumnbconvert has an Arbitrary File Read via Path Traversal in HTMLExporter Image Embeddingnbconvert
- HighClaude Code: Sandbox Escape via Symlink Following Allows Arbitrary File Write Outside Workspace@anthropic-ai/claude-code
- LowLangflow has an Information Leak through Incomplete API Key Redactionlangflow
- MediumAgentScope vulnerable to Server-Side Request Forgeryagentscope
- MediumAgentScope vulnerable to Server-Side Request Forgeryagentscope
- LowLangflow: Cleartext Storage of Authentication Settings in Project Creation Endpointlangflow
- MediumAgentScope Vulnerable to Remote Code Injectionagentscope
- MediumAgentScope vulnerable to Server-Side Request Forgeryagentscope
- MediumApache Doris MCP Server vulnerable to SQL Injection via improper query context neutralizationdoris-mcp-server
- LowLangflow vulnerable to injectionlangflow
- MediumLangflow: DoS Through Lack of File Size Restriction via Deprecated Unauthenticated File Upload APIlangflow-base
- CriticalFlowise: Airtable_Agent Code Injection Remote Code Execution Vulnerabilityflowise, flowise-components
- MediumOpenClaw: QQBot reply media URL handling could trigger SSRF and re-upload fetched bytesopenclaw
- HighOpenClaw: busybox and toybox applet execution weakened exec approval bindingopenclaw
- HighOpenClaw: Matrix room control-command authorization no longer trusts DM pairing-store entriesopenclaw
- HighFlowise: Mass Assignment in DocumentStore Create Endpoint Leads to Cross-Workspace Object Takeover (IDOR)flowise
- MediumOpenClaw: Empty approver lists could grant explicit approval authorizationopenclaw
- HighOpenClaw: Sandbox browser CDP relay could expose DevTools protocol on 0.0.0.0openclaw
- MediumOpenClaw: Existing-session browser interaction routes bypassed SSRF policy enforcementopenclaw
- MediumOpenClaw: Browser press/type interaction routes missed complete navigation guard coverageopenclaw
- MediumOpenClaw: Browser SSRF policy default allowed private-network navigationopenclaw
- MediumClaude Code: Insecure System-Wide Configuration Loading Enables Local Privilege Escalation on Windows@anthropic-ai/claude-code
- HighFlowise: Unauthenticated TTS endpoint accepts arbitrary credential IDs — enables API credit abuse via stored credentialsflowise
- MediumOpenClaw: Memory dreaming config persistence was reachable from operator.write commandsopenclaw
- HighOpenClaw: QQBot media tags could read arbitrary local files through reply textopenclaw
- HighOpenClaw: Sandboxed agents could escape exec routing via host=node overrideopenclaw
- MediumOpenClaw: Agent hook events could enqueue trusted system events from unsanitized external inputopenclaw
- HighOpenClaw: Matrix profile config persistence was reachable from operator.write message toolsopenclaw
- MediumOpenClaw: Workspace .env could inject OpenClaw runtime-control variablesopenclaw
- HighOpenClaw: Channel setup catalog lookups could include untrusted workspace plugin shadowsopenclaw
- HighOpenClaw: config.get redaction bypass through sourceConfig and runtimeConfig aliasesopenclaw
- MediumOpenClaw: Sandbox noVNC helper route exposed interactive browser session credentialsopenclaw
- HighOpenClaw: Workspace provider auth choices could auto-enable untrusted provider pluginsopenclaw
- CriticalPraisonAI has an incomplete fix for CVE-2026-34935 - OS Command Injectionpraisonai
- MediumOpenClaw: Browser snapshot and screenshot routes could expose internal page content after navigationopenclaw
- MediumOpenClaw: Discord event cover images bypassed sandbox media normalizationopenclaw
- MediumOpenClaw: Nostr profile mutation routes allowed operator.write config persistenceopenclaw
- MediumOpenClaw: CDP /json/version WebSocket URL could pivot to untrusted second-hop targetsopenclaw
- MediumOpenClaw: QMD memory_get restricts reads to canonical or indexed memory pathsopenclaw
- MediumOpenClaw: Heartbeat owner downgrade missed untrusted webhook wake eventsopenclaw
- MediumOpenClaw: Heartbeat owner downgrade missed local async exec completion eventsopenclaw
- LowOpenClaw: Microsoft Teams SSO invoke handler missed sender authorization checksopenclaw
- MediumOpenClaw: Shell-wrapper detection missed env-argv assignment injection formsopenclaw
- MediumOpenClaw: screen_record outPath bypassed workspace-only filesystem guardopenclaw
- MediumOpenClaw: Sender policy bypass in host media attachment reads allows unauthorized local file disclosureopenclaw
- MediumOpenClaw: Collect-mode queue batches could reuse the last sender authorization contextopenclaw
- MediumOpenClaw: Webchat media embedding enforces local-root containment for tool-result filesopenclaw
- MediumOpenClaw: Browser interaction routes could pivot into local CDP and regain file readsopenclaw
- LowOpenClaw: Delivery queue recovery could lose group tool-policy context for media replayopenclaw
- HighPraisonAI: SQL Injection via unvalidated `table_prefix` in 9 conversation store backends (incomplete fix for CVE-2026-40315)praisonaiagents, praisonai
- MediumOpenClaw: Browser tabs action select and close routes bypassed SSRF policyopenclaw
- HighOpenClaw: Exec environment denylist missed high-risk interpreter startup variablesopenclaw
- HighOpenClaw: Voice-call realtime WebSocket accepted oversized framesopenclaw
- HighFlowise: Public chatflow endpoints return unsanitized flowData including plaintext API keys, passwords, and credential IDsflowise
- LowNeo4j Labs MCP Servers: SSRF and Data Modification via read_only Mode Bypass Through CALL Proceduresmcp-neo4j-cypher
- CriticalOpenClaw: Feishu webhook and card-action validation now fail closedopenclaw
- CriticalOpenClaw: Gateway HTTP endpoints re-resolve bearer auth after SecretRef rotationopenclaw
- MediumOpenClaw: Browser SSRF hostname validation could be bypassed by DNS rebindingopenclaw
- HighFlowise: Cypher Injection in GraphCypherQAChainflowise, flowise-components
- MediumFlowise: Weak Default Express Session Secretflowise
- HighFlowise: SSRF Protection Bypass (TOCTOU & Default Insecure)flowise, flowise-components
- HighMCP-Framework: Unbounded memory allocation in readRequestBody allows denial of service via HTTP transportmcp-framework
- HighFlowise: Improper Mass Assignment in Account Registration Enables Unauthorized Organization Associationflowise
- HighFlowise: Sensitive Data Leak in public-chatbotConfigflowise
- HighFlowise: Unauthenticated OAuth 2.0 Access Token Disclosure via Public Chatflow in Flowiseflowise
- MediumFlowise: Unauthenticated Information Disclosure of OAuth Secrets (Cleartext) via GET Requestflowise
- HighFlowise: APIChain Prompt Injection SSRF in GET/POST API Chainsflowise, flowise-components
- MediumFlowise Execute Flow function has an SSRF vulnerabilityflowise, flowise-components
- CriticalFlowise: Code Injection in CSVAgent leads to Authenticated RCEflowise, flowise-components
- CriticalFlowise: Authenticated RCE Via MCP Adaptersflowise, flowise-components
- MediumFlowise: Weak Default JWT Secretsflowise
- HighFlowise: Parameter Override Bypass Remote Command Executionflowise, flowise-components
- HighFlowise: Remote code execution vulnerability in AirtableAgent.ts caused by lack of input verification when using `Pandas`.flowise, flowise-components
- HighFlowise: resetPassword Authentication Bypass Vulnerabilityflowise
- LowOpenClaw: TOCTOU read in exec script preflightopenclaw
- MediumFlowise: Weak Default Token Hash Secretflowise
- MediumFlowise: SSRF Protection Bypass via Direct node-fetch / axios Usage (Patch Enforcement Failure)flowise, flowise-components
- HighFlowise: File Upload Validation Bypass in createAttachmentflowise
- MediumLangSmith SDK: Streaming token events bypass output redactionlangsmith
- MediumFlowise: Path Traversal in Vector Store basePathflowise, flowise-components
- HighFlowise: Password Reset Link Sent Over Unsecured HTTPflowise
- HighFlowise: SSRF Protection Bypass via Unprotected Built-in HTTP Modules in Custom Function Sandboxflowise, flowise-components
- HighMCP Server Kubernetes has an Argument Injection in port_forward tool via space-splittingmcp-server-kubernetes
- Criticalexcel-mcp-server has a Path Traversal issueexcel-mcp-server
- HighSSH/SCP option injection allowing local RCE in @aiondadotcom/mcp-ssh@aiondadotcom/mcp-ssh
- CriticalOpenAI Codex CLI enables code execution through malicious MCP (Model Context Protocol) configuration files@openai/codex
- HighKeras has an untrusted deserialization vulnerabilitykeras
- HighApache SkyWalking MCP: Server-Side Request Forgery via SW-URL Header in MCP Servergithub.com/apache/skywalking-mcp
- CriticalGoogle Agent Development Kit (ADK) has a Code Injection and Missing Authentication vulnerabilitygoogle-adk, google-adk
- LowMetaGPT affected by server-side request forgery in metagpt/utils/common.pymetagpt
- LowMetaGPT has an eval injection via a cross-site request forgery attackmetagpt
- MediumMetaGPT has an eval injection in metagpt/strategy/tot.pymetagpt
- Criticalaws-mcp has a Command Injection Remote Code Execution Vulnerabilityaws-mcp
- HighPraisonAI Vulnerable to Implicit Execution of Arbitrary Code via Automatic `tools.py` Loadingpraisonai
- MediumPraisonAI has Unrestricted Upload Size in WSGI Recipe Registry Server that Enables Memory Exhaustion DoSpraisonai
- HighPraisonAI Vulnerable to Code Injection and Protection Mechanism Failurepraisonai
- HighPraisonAI: Unauthenticated Allow-List Manipulation Bypasses Agent Tool Approval Safety Controlspraisonai
- LowOpenClaw vulnerable to SSRF in src/agents/tools/web-fetch.tsopenclaw
- Highn8n-mcp has unauthenticated session termination and information disclosure in HTTP transportn8n-mcp
- MediumPraisonAIAgents: Path Traversal via Unvalidated Glob Pattern in list_files Bypasses Workspace Boundarypraisonaiagents
- HighPraisonAIAgents has SSRF and Local File Read via Unvalidated URLs in web_crawl Toolpraisonaiagents
- HighPraisonAI Vulnerable to Server-Side Request Forgery via Unvalidated webhook_url in Jobs APIpraisonai
- CriticalPraisonAI Browser Server allows unauthenticated WebSocket clients to hijack connected extension sessionspraisonaiagents, praisonai
- CriticalPraisonAI vulnerable to arbitrary file write via path traversal in `praisonai recipe unpack`praisonai
- MediumPraisonAI Vulnerable to Stored XSS via Unsanitized Agent Output in HTML Rendering (nh3 Not a Required Dependency)praisonai
- MediumPraisonAI Vulnerable to Decompression Bomb DoS via Recipe Bundle Extraction Without Size Limitspraisonai
- MediumPraisonAI: Coarse-Grained Tool Approval Cache Bypasses Per-Invocation Consent for Shell Commandspraisonaiagents
- HighPraisonAI Vulnerable to Argument Injection into Cloud Run Environment Variables via Unsanitized Comma in gcloud --set-env-varspraisonai
- HighPraisonAI Vulnerable to RCE via Automatic tools.py Importpraisonaiagents, praisonai
- MediumPraisonAIAgents: Arbitrary File Read via read_skill_file Missing Workspace Boundary and Approval Gatepraisonaiagents
- MediumPraisonAI Vulnerable to Sensitive Environment Variable Exposure via Untrusted MCP Subprocess Executionpraisonai
- MediumPraisonAI: Unauthenticated Information Disclosure of Agent Instructions via /api/agents in AgentOSpraisonai
- CriticalPraisonAI Vulnerable Untrusted Remote Template Code Executionpraisonai
- HighPraisonAI: Unauthenticated WebSocket Endpoint Proxies to Paid OpenAI Realtime API Without Rate Limitspraisonai
- HighPraisonAIAgents: SSRF via unvalidated URL in `web_crawl` httpx fallbackpraisonaiagents
- HighPraisonAI: Hardcoded `approval_mode="auto"` in Chainlit UI Overrides Administrator Configuration, Enabling Unapproved Shell Command Executiopraisonai
- CriticalPraisonAIAgents has an OS Command Injection via shell=True in Memory Hooks Executor (memory/hooks.py)praisonaiagents
- HighPraisonAIAgents: Environment Variable Secret Exfiltration via os.path.expandvars() Bypassing shell=False in Shell Toolpraisonaiagents
- CriticalPraisonAI has critical RCE via `type: job` workflow YAMLpraisonaiagents, praisonai
- HighPraisonAI: Cross-Origin Agent Execution via Hardcoded Wildcard CORS and Missing Authentication on AGUI Endpointpraisonaiagents
- MediumPraisonAI: SQLiteConversationStore didn't validate table_prefix when constructing SQL queriespraisonai
- LowOpenClaw: Concurrent async auth attempts can bypass the intended shared-secret rate-limit budget on Tailscale-capable pathsopenclaw
- MediumOpenClaw QQ Bot Extension missing SSRF Protection on All Media Fetch Pathsopenclaw
- MediumFoundationAgents MetaGPT vulnerable to eval injectionmetagpt
- MediumOpenClaw B-M3: ClawHub package downloads are not enforced with integrity verificationopenclaw
- LowOpenClaw: Gateway plugin HTTP `auth: gateway` widens identity-bearing `operator.read` requests into runtime `operator.write`openclaw
- LowOpenClaw: Feishu docx upload_file/upload_image Bypasses Workspace-Only Filesystem Policy (GHSA-qf48-qfv4-jjm9 Incomplete Fix)openclaw
- MediumOpenClaw: Existing WS sessions survive shared gateway token rotationopenclaw
- HighOpenClaw: Node Pairing Reconnect Command Escalation Bypasses operator.admin Scope Requirementopenclaw
- MediumOpenClaw `node.pair.approve` placed in `operator.write` scope instead of `operator.pairing` allows unprivileged pairing approvalopenclaw
- MediumOpenClaw: resolvedAuth closure becomes stale after config reloadopenclaw
- HighOpenClaw: HGRCPATH, CARGO_BUILD_RUSTC_WRAPPER, RUSTC_WRAPPER, and MAKEFLAGS missing from exec env denylist — RCE via build tool env injectioopenclaw
- Lowawwaiid mcp-server-taskwarrior vulnerable to command injectionmcp-server-taskwarrior
- MediumOpenClaw: Multiple Code Paths Missing Base64 Pre-Allocation Size Checksopenclaw
- LowOpenClaw: GIT_DIR and related git plumbing env vars missing from exec env denylist (GHSA-m866-6qv5-p2fg variant)openclaw
- MediumOpenClaw `node.invoke(browser.proxy)` bypasses `browser.request` persistent profile-mutation guardopenclaw
- Mediumapi-lab-mcp vulnerable to SSRFapi-lab-mcp
- MediumFoundationAgents MetaGPT vulnerable to OS Command Injection in metagpt/tools/libs/terminal.pymetagpt
- MediumMetaGPT has an Injection issuemetagpt
- HighOpenClaw: Lower-trust background runtime output is injected into trusted `System:` events, and local async exec completion misses the intendopenclaw
- HighOpenClaw: Authenticated `/hooks/wake` and mapped `wake` payloads are promoted into the trusted `System:` prompt channelopenclaw
- MediumOpenClaw: strictInlineEval explicit-approval boundary bypassed by approval-timeout fallback on gateway and node exec hostsopenclaw
- MediumOpenClaw: Shared reply MEDIA - paths are treated as trusted and can trigger cross-channel local file exfiltrationopenclaw
- MediumFoundationAgents MetaGPT vulnerable to OS Command Injection in metagpt/utils/common.pymetagpt
- HighOpenClaw: `fetchWithSsrFGuard` replays unsafe request bodies across cross-origin redirectsopenclaw
- MediumOpenClaw: /allowlist omits owner-only enforcement for cross-channel allowlist writesopenclaw
- MediumOpenClaw has Browser SSRF Policy Bypass via Interaction-Triggered Navigationopenclaw
- MediumOpenClaw: Strict browser SSRF bypass in Playwright redirect handling leaves private targets reachableopenclaw
- MediumOpenClaw Host-Exec Environment Variable Injectionopenclaw
- MediumOpenClaw `device.token.rotate` mints tokens for unapproved roles, bypassing device role-upgrade pairingopenclaw
- MediumFoundationAgents MetaGPT vulnerable to os command injection via the Terminal.run_commandmetagpt
- CriticalMarimo: Pre-Auth Remote Code Execution via Terminal WebSocket Authentication Bypassmarimo
- CriticalPraisonAI Vulnerable to OS Command Injectionpraisonai
- CriticalPraisonAI Vulnerable to Remote Code Execution via YAML Deserialization in Agent Definition Loadingpraisonai
- Highn8n-mcp has authenticated SSRF via instance-URL header in multi-tenant HTTP moden8n-mcp
- HighLiteLLM: Password hash exposure and pass-the-hash authentication bypasslitellm
- MediumPraisonAI has Memory State Leakage and Path Traversal in MultiAgent Context Handlingpraisonaiagents
- MediumLangChain has incomplete f-string validation in prompt templateslangchain-core, langchain-core
- HighPraisonAI Has Unauthenticated SSE Event Stream that Exposes All Agent Activity in A2U Serverpraisonai
- HighPraisonAI has Template Injection in Agent Tool Definitionspraisonai
- Highstata-mcp has insufficient validation of user-supplied Stata do-file content that can lead to command executionstata-mcp
- CriticalPraisonAI has sandbox escape via exception frame traversal in `execute_code` (subprocess mode)praisonaiagents
- Highmcp-from-openapi is Vulnerable to SSRF via $ref Dereferencing in Untrusted OpenAPI Specificationsmcp-from-openapi
- MediumOpenClaw: Gateway hello snapshots exposed host config and state paths to non-admin clientsopenclaw
- MediumOpenClaw: Untrusted workspace channel shadows could execute during built-in channel setupopenclaw
- MediumOpenClaw: Gateway operator.write Can Reach Admin-Class Talk Voice Config Persistence via chat.sendopenclaw
- MediumOpenClaw: OpenShell `mirror` mode can convert untrusted sandbox files into explicitly enabled workspace hooks and execute them on the host dopenclaw
- MediumMLflow is vulnerable to an authorization bypass affecting the AJAX endpointmlflow
- MediumOpenClaw: Tlon media downloads can bypass core safety limits and exhaust diskopenclaw
- MediumOpenClaw: iOS A2UI bridge trusted generic local-network pages for agent.request dispatchopenclaw
- MediumOpenClaw: Read-scoped identity-bearing HTTP clients could kill sessions via /sessions/:sessionKey/killopenclaw
- MediumHuggingFace Transformers allows for arbitrary code execution in the `Trainer` classtransformers
- HighOpenClaw: Gateway operator.write Can Reach Admin-Class Telegram Config and Cron Persistence via sendopenclaw
- MediumOpenClaw: Android accepted cleartext remote gateway endpoints and sent stored credentials over ws://openclaw
- MediumOpenClaw: QQ Bot structured payloads could read arbitrary local filesopenclaw
- HighJava-SDK has a DNS Rebinding Vulnerabilityio.modelcontextprotocol.sdk:mcp-core
- MediumOpenClaw: Windows-compatible env override keys could bypass system.run approval bindingopenclaw
- MediumOpenClaw: Trailing-dot localhost CDP hosts could bypass remote loopback protectionsopenclaw
- MediumMLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interfacemlflow
- LowOpenClaw: Zalo replay dedupe cache could suppress events across authenticated webhook targetsopenclaw
- MediumOpenClaw: `session_status` still bypasses configured `tools.sessions.visibility` for unsandboxed invocationsopenclaw
- MediumOpenClaw: `/phone arm`/`/phone disarm` Bypasses `operator.admin` Scope Check for External Channelsopenclaw
- MediumOpenClaw: Forged Nostr DMs could create pairing state before signature verificationopenclaw
- MediumOpenClaw: Shared-secret comparison call sites leaked length information through timingopenclaw
- MediumOpenClaw: OpenShell mirror mode could delete arbitrary remote directories when roots were mis-scopedopenclaw
- MediumOpenClaw: Zalo replay dedupe keys could suppress messages across chats or sendersopenclaw
- HighOpenClaw Has Incomplete Fix for CVE-2026-4039: CLI Backend Environment Variable Injection via Workspace Configopenclaw
- MediumOpenClaw: Marketplace Plugin Download Follows Redirects Without SSRF Protectionopenclaw
- MediumOpenClaw: pnpm dlx approvals did not bind local script operandsopenclaw
- MediumOpenClaw: Shell init-file options could satisfy exec allowlist script matchingopenclaw
- MediumOpenClaw: Pairing pending-request caps were enforced per channel instead of per accountopenclaw
- HighPraisonAI Has Arbitrary File Write (Zip Slip) in Templates Extractionpraisonai
- HighPraisonAI recipe registry pull path traversal writes files outside the chosen output directorypraisonai
- CriticalPraisonAI Has Path Traversal in FileToolspraisonai
- MediumOpenClaw's complex interpreter pipelines could skip exec script preflight validationopenclaw
- CriticalPraisonAI Vulnerable to Arbitrary File Write / Path Traversal in Action Orchestratorpraisonai
- HighPraisonAI recipe registry publish path traversal allows out-of-root file writepraisonai
- Low@nor2/heim-mcp vulnerable to command injection@nor2/heim-mcp
- Low@elgentos/magento2-dev-mcp vulnerable to command injection@elgentos/magento2-dev-mcp
- High@mobilenext/mobile-mcp: Arbitrary Android Intent Execution via mobile_open_url@mobilenext/mobile-mcp
- HighOpenClaw: Gemini OAuth exposed the PKCE verifier through the OAuth state parameteropenclaw
- MediumOpenClaw: Voice-call still parses large WebSocket frames before start validation (Incomplete fix for CVE-2026-32062)openclaw
- MediumOpenClaw: Telnyx Webhook Replay Detection Bypass via Base64 Signature Re-encodingopenclaw
- MediumvLLM: Unauthenticated OOM Denial of Service via Unbounded `n` Parameter in OpenAI API Servervllm
- LowOpenClaw: Tlon Startup Migration Rehydrates Empty-Array Revocations From File Configopenclaw
- MediumJupyterHub has an Open Redirect Vulnerabilityjupyterhub
- MediumOpenClaw: diffs viewer misclassifies proxied remote requests as loopback when `allowRemoteViewer` is disabledopenclaw
- HighLiteLLM: Privilege escalation via unrestricted proxy configuration endpointlitellm
- HighOpenClaw: Self-Whitelisting in appendLocalMediaParentRoots Allows Arbitrary File Read & Credential Exfiltrationopenclaw
- MediumOpenClaw: Path traversal via inbound channel attachment path in ACP dispatch allows arbitrary file readopenclaw
- LowOpenClaw: Discord Component Interaction Misclassifies Group DM as Direct Messageopenclaw
- MediumOpenClaw: Media download follows cross-origin redirects with Authorization headers intactopenclaw
- MediumOpenClaw: Fake DeviceToken Bypasses Shared Auth Rate Limitingopenclaw
- Criticalmlflow: FastAPI job endpoints under `/ajax-api/3.0/jobs/*` are not protected by authentication or authorizationmlflow
- MediumOpenClaw: Endpoint persists after trust decline, leaking gateway credentialsopenclaw
- MediumOpenClaw: Host exec environment overrides miss proxy, TLS, Docker, and Git TLS controlsopenclaw
- CriticalOpenClaw: Sandbox escape via TOCTOU race in remote FS bridge readFileopenclaw
- HighOpenClaw: Host exec environment sanitization misses package, registry, Docker, compiler, and TLS override variablesopenclaw
- MediumOpenClaw: Discord voice manager bypasses channel-level member access allowlistopenclaw
- HighOpenClaw: OpenShell Mirror Sync — Sandbox Escape via Unrestricted File Sync + Symlink Traversalopenclaw
- MediumOpenClaw: Telegram legacy allowFrom migration fans default-account trust into all named accountsopenclaw
- HighOpenClaw: Media Parsing Path Traversal Leads to Arbitrary File Readopenclaw
- HighBentoML: Command Injection in cloud deployment setup scriptbentoml
- HighOpenClaw: Incomplete scope-clearing fix allows operator.admin escalation via trusted-proxy auth modeopenclaw
- HighOpenClaw: Incomplete host-env-security-policy allows untrusted model to substitute compiler binaries via env overridesopenclaw
- HighOpenClaw: Unbound bootstrap setup codes allow privilege escalation during pairingopenclaw
- HighOpenClaw: Paired node escalates to gateway RCE via unrestricted node.event agent dispatchopenclaw
- HighOpenClaw: Node browser proxy `allowProfiles` bypass through persistent profile mutation and runtime profile selectionopenclaw
- MediumOpenClaw runs Discord audio preflight transcription before member authorizationopenclaw
- MediumOpenClaw Has a Gateway Control Interface Information Disclosure Vulnerabilityopenclaw
- CriticalLiteLLM: Authentication bypass via OIDC userinfo cache key collisionlitellm
- MediumOpenClaw: Telegram audio preflight transcription enables resource consumption by unauthorized sendersopenclaw
- LowOpenClaw: HTTP operator endpoints lack browser-origin validation in trusted-proxy modeopenclaw
- HighOpenClaw: MS Teams webhook parses body before JWT validation, enabling unauthenticated resource exhaustionopenclaw
- MediumvLLM: Server-Side Request Forgery (SSRF) in `download_bytes_from_url `vllm
- MediumvLLM: Denial of Service via Unbounded Frame Count in video/jpeg Base64 Processingvllm
- HighOpenClaw: macOS Tailnet DNS Spoofing & Credential Exfiltrationopenclaw
- HighOpenClaw: Workspace `.env` can override the bundled plugin trust rootopenclaw
- LowOpenClaw: Gateway `device.token.rotate` does not terminate active WebSocket sessions after credential rotationopenclaw
- MediumOpenClaw: Sandbox file operations use check-then-act, bypassing fd-based TOCTOU defensesopenclaw
- LowOpenClaw: Discord Slash Commands Bypass Group DM Channel Allowlistopenclaw
- HighOpenClaw: Agentic Consent Bypass — LLM Agent Can Silently Disable Exec Approval via `config.patch`openclaw
- HighBentoML: SSTI via Unsandboxed Jinja2 in Dockerfile Generationbentoml
- MediumOpenClaw: Image pixel-limit guard can fail open on sips and allow decompression-bomb DoSopenclaw
- LowOpenClaw: Discord voice ingress authorization can be bypassed via channel, name, and stale-role validation gapsopenclaw
- HighOpenClaw: Device-Paired Node Skips Node Scope Gate → Host RCE.mdopenclaw
- HighOpenClaw: Workspace `.env` can override the bundled hooks root and load attacker hook codeopenclaw
- Lowfast-filesystem-mcp is vulnerable to command injection through handleGetDiskUsage functionfast-filesystem-mcp
- HighOpenClaw: PIP_INDEX_URL and UV_INDEX_URL bypass host exec env sanitization and redirect Python package-index trafficopenclaw
- MediumOpenClaw: Feishu thread history and quoted messages bypass sender allowlistopenclaw
- MediumOpenClaw: Voice-call Plivo replay mutates in-process callback origin before replay rejectionopenclaw
- MediumOpenClaw: SSRF via Unguarded `fetch()` in Marketplace Plugin Download and Ollama Model Discoveryopenclaw
- LowOpenClaw: MSTeams thread history bypasses sender allowlist via Graph APIopenclaw
- LowOpenClaw: Security Scan Failure Does Not Block Plugin Installation (Fail-Open)openclaw
- HighOpenClaw: SSH sandbox tar upload follows symlinks, enabling arbitrary file write on remote hostopenclaw
- CriticalOpenClaw: Heartbeat context inheritance bypasses sandbox via senderIsOwner escalationopenclaw
- LowOpenClaw: Zalo webhook replay cache cross-target messageId scope bypassopenclaw
- MediumOpenClaw: SSH-based sandbox backends pass unsanitized process.env to child processesopenclaw
- MediumOpenClaw Nostr privateKey config redaction bypass leaks plaintext signing key via config.getopenclaw
- MediumOpenClaw: Unauthenticated plugin-auth HTTP routes receive operator runtime scopesopenclaw
- Lowa11y-mcp: Server-Side Request Forgery (SSRF) vulnerability in A11yServer functiona11y-mcp
- MediumOpenClaw: LINE webhook handler lacks shared pre-auth concurrency budget before signature verificationopenclaw
- LowOpenClaw: Matrix thread root and reply context bypass sender allowlistopenclaw
- MediumOpenClaw: node.pair.approve missing callerScopes validation allows low-privilege operator to approve malicious nodesopenclaw
- HighPraisonAI Has SSRF in FileTools.download_file() via Unvalidated URLpraisonaiagents
- HighONNX: Malicious ONNX models can crash servers by exploiting unprotected object settings.onnx
- MediumClaude SDK for TypeScript: Memory Tool Path Validation Allows Sandbox Escape to Sibling Directories@anthropic-ai/sdk
- HighOpenClaw Gateway `operator.write` can reach admin-only session reset via `chat.send` `/reset`openclaw
- MediumOpenClaw: Google Chat and Zalouser group sender allowlist bypass via policy downgradeopenclaw
- CriticalPraisonAI: Python Sandbox Escape via str Subclass startswith() Override in execute_codepraisonaiagents
- HighOpen WebUI has Broken Access Control in Tool Valvesopen-webui
- CriticalOpenClaw has a CWD `.env` environment variable injection which bypasses host-env policy and allows config takeoveropenclaw
- MediumPraisonAI Has ReDoS via Unvalidated User-Controlled Regex in MCPToolIndex.search_tools()praisonai
- CriticalPraisonAI Has Authentication Bypass via OAuthManager.validate_token()praisonai
- CriticalPraisonAI Has Second-Order SQL Injection in `get_all_user_threads`praisonai
- CriticalPraisonAI: OS Command Injection in MCPHandler.parse_mcp_command()praisonai
- CriticalPraisonAI Has Missing Authentication in WebSocket Gatewaypraisonai
- MediumONNX: Arbitrary File Read via ExternalData Hardlink Bypass in ONNX loadonnx
- CriticalMalware in mcp-server-todomcp-server-todo
- MediumONNX: External Data Symlink Traversalonnx
- HighOpenClaw gateway exec allow-always over-trusts positional carrier executablesopenclaw
- HighONNX: TOCTOU arbitrary file read/write in save_external_datonnx
- MediumClaude SDK for Python has Insecure Default File Permissions in Local Filesystem Memory Toolanthropic
- LowOpenClaw affected by SSRF via unguarded image download in fal provideropenclaw
- HighPraisonAI Has Sandbox Escape via shell=True and Bypassable Blocklist in SubprocessSandboxpraisonai
- Highmcp-handler has a tool response leak across concurrent client sessions ('Race Condition')mcp-handler
- HighPraisonAI: Shell Injection in run_python() via Unescaped $() Substitutionpraisonaiagents
- MediumClaude SDK for Python: Memory Tool Path Validation Race Condition Allows Sandbox Escapeanthropic
- HighPraisonAI: SSRF via Unvalidated api_base in passthrough() Fallbackpraisonai
- HighDNS Rebinding Protection Disabled by Default in Model Context Protocol Go SDK for Servers Running on Localhostgithub.com/modelcontextprotocol/go-sdk
Mar 2026366 advisories · 1 malicious packages · 1 exploited
- HighOpenClaw's device removal and token revocation do not terminate active WebSocket sessionsopenclaw
- HighOpenClaw: CLI Remote Onboarding Persists Unauthenticated Discovery Endpoint and Exfiltrates Gateway Credentialsopenclaw
- Highonnx Vulnerable to Path Traversal via Symlinkonnx
- MediumOpenClaw: Gateway `operator.write` can reach admin-only persisted `verboseLevel` via `chat.send` `/verbose`openclaw
- HighOpenClaw has a gateway exec allowlist allow-always bypass via unregistered /usr/bin/script wrapperopenclaw
- HighOpenClaw: Gateway chat.send ACP-only provenance guard could be bypassed by client identity spoofingopenclaw
- HighOpenClaw: Voice-call Plivo V3 webhook replay key uses unsorted URL, allowing replay via query-parameter reorderingopenclaw
- MediumOpenClaw's Nextcloud Talk webhook missing rate limiting on shared secret authenticationopenclaw
- HighOpenClaw: Discord text `/approve` bypasses `channels.discord.execApprovals.approvers` and allows non-approvers to resolve pending exec approopenclaw
- MediumOpenClaw: Gateway WebSocket Denial of Service via unbounded pre-auth upgradesopenclaw
- LowOpenClaw SSRF guard misses four IPv6 special-use rangesopenclaw
- CriticalOpenClaw: /pair approve command path omitted caller scope subsetting and reopened device pairing escalationopenclaw
- HighOpenClaw's incomplete host env sanitization blocklist allows supply-chain redirection via package-manager env overridesopenclaw
- HighOpenClaw safeBins jq `$ENV` filter bypass allows environment variable disclosureopenclaw
- MediumOpenClaw's Discord component interaction ingress skips guild/channel policy enforcementopenclaw
- MediumOpenClaw host-env blocklist missing `GIT_TEMPLATE_DIR` and `AWS_CONFIG_FILE` allows code execution via env overrideopenclaw
- MediumFastMCP has a Command Injection vulnerability - Gemini CLIfastmcp
- MediumOpenClaw: Feishu extension resolveUploadInput bypasses file-system sandbox and allows arbitrary file reads via upload_imageopenclaw
- CriticalMlflow: Command Injection when serving models with enable_mlserver=Truemlflow
- HighFastMCP: Missing Consent Verification in OAuth Proxy Callback Facilitates Confused Deputy Vulnerabilitiesfastmcp
- MediumOpenClaw: Zalo channel downloads media before sender authorizationopenclaw
- HighOpenClaw's message tool media parameter bypasses tool policy filesystem isolationopenclaw
- CriticalFastMCP OpenAPI Provider has an SSRF & Path Traversal Vulnerabilityfastmcp
- MediumOpenClaw has incomplete Fix for CVE-2026-27486: Unvalidated SIGKILL in `!stop` Chat Command via `shell-utils.ts`openclaw
- MediumOpenClaw: Non-owner command-authorized sender can change the owner-only `/send` session delivery policyopenclaw
- MediumOpenClaw has a Gateway HTTP /v1/models Route Bypasses Operator Read Scopeopenclaw
- HighOpenClaw: Gateway operator.write Can Reach Admin-Class Channel Allowlist Persistence via chat.sendopenclaw
- HighOpenClaw has Sandbox Media Root Bypass via Unnormalized `mediaUrl` / `fileUrl` Parameter Keys (CWE-22)openclaw
- MediumMCP Java SDK has a Hardcoded Wildcard CORS (Access-Control-Allow-Origin: *)io.modelcontextprotocol.sdk:mcp-core, io.modelcontextprotocol.sdk:mcp-core, io.modelcontextprotocol.sdk:mcp-core
- HighOpenClaw has an Arbitrary Malicious Code Execution Vulnerabilityopenclaw
- CriticalMLflow Command Injection vulnerabilitymlflow
- CriticalMLFlow path traversal vulnerabilitymlflow
- MediumOpenClaw: Mutating internal `/allowlist` chat commands missed `operator.admin` scope enforcementopenclaw
- MediumOpenClaw has incomplete Fix for CVE-2026-32011: Feishu Webhook Pre-Auth Body Parsing DoS (Slow-Body / Slowloris Variant)openclaw
- HighOpenClaw: `browser.request` still allows `POST /reset-profile` through the `operator.write` surfaceopenclaw
- MediumOpenClaw: Feishu webhook reads and parses unauthenticated request bodies before signature validationopenclaw
- MediumOpenClaw has ACP CLI approval prompt ANSI escape sequence injectionopenclaw
- LowOpenClaw: Google Chat Authz Bypass via Group Policy Rebinding with Mutable Space displayNameopenclaw
- MediumOpenClaw: Gateway HTTP Session History Route Bypasses Operator Read Scopeopenclaw
- MediumOpenClaw: Feishu Raw Card Send Surface Can Mint Legacy Card Callbacks That Bypass DM Pairingopenclaw
- MediumOpenClaw: Gateway Plugin Subagent Fallback `deleteSession` Uses Synthetic `operator.admin`openclaw
- MediumOpenClaw: Telegram DM-Scoped Inline Button Callbacks Bypass DM Pairing and Mutate Session Stateopenclaw
- MediumOpenClaw: Synology Chat Webhook Pre-Auth Rate-Limit Bypass Enables Brute-Force Guessing of Webhook Tokenopenclaw
- HighOpenClaw: `session_status` sessionId resolution bypasses sandboxed session-tree visibilityopenclaw
- MediumOpenClaw: MS Teams Feedback Invocation Bypasses Sender Allowlists and Records Unauthorized Session Feedbackopenclaw
- HighOpenClaw: SSRF via Unguarded Configured Base URLs in Multiple Channel Extensions (Incomplete Fix for CVE-2026-28476)openclaw
- MediumOpen WebUI has unauthorized deletion of knowledge filesopen-webui
- Mediumn8n has XSS in its Credential Management Flown8n, n8n
- Mediumn8n has XSS in Chat Trigger Node through Custom CSSn8n, n8n, n8n
- High@mobilenext/mobile-mcp alllows arbitrary file write via Path Traversal in mobile screen capture tools@mobilenext/mobile-mcp
- LowHugging Face Smolagents has an Injection issuesmolagents
- HighvLLM has Hardcoded Trust Override in Model Files Enables RCE Despite Explicit User Opt-Outvllm
- HighLangflow: Authenticated Users Can Read, Modify, and Delete Any Flow via Missing Ownership Checklangflow, langflow-base
- CriticalOpenClaw: Gateway Backend Reconnect lets Non-Admin Operator Scopes Self-Claim operator.adminopenclaw
- HighOpenClaw: Gateway HTTP /sessions/:sessionKey/kill Reaches Admin Kill Path Without Caller Scope Bindingopenclaw
- MediumOpenClaw: Matrix Verification Notices Bypass Matrix DM Policy and Reply to Unpaired DM Peersopenclaw
- CriticalOpenClaw: Silent privilege escalation via gateway shared-auth reconnectopenclaw
- CriticalSpring AI: SpEL injection is triggered when a user-supplied value is used as a filter expression keyorg.springframework.ai:spring-ai-vector-store, org.springframework.ai:spring-ai-vector-store
- HighMLFlow allows Tracing + Assessments Accessmlflow
- HighOpen WebUI's process_files_batch() endpoint missing ownership check, allows unauthorized file overwriteopen-webui
- MediumOpenClaw: BlueBubbles Group Reactions Bypass requireMention and Still Enqueue Agent-Visible System Eventsopenclaw
- Mediumn8n has a Stored XSS Vulnerability in its Form Triggern8n, n8n
- HighLangChain Core has Path Traversal vulnerabilites in legacy `load_prompt` functionslangchain-core
- HighOpenClaw: Gateway Plugin HTTP Auth Grants Unrestricted operator.admin Runtime Scope to All Callersopenclaw
- HighMCP Ruby SDK: Insufficient Session Binding Allows SSE Stream Hijacking via Session ID Replaymcp
- MediumOpenClaw: Telegram Webhook Missing Guess Rate Limiting Enables Brute-Force Guessing of Weak Webhook Secretopenclaw
- HighAzure Data Explorer MCP Server: KQL Injection in multiple tools allows MCP client to execute arbitrary Kusto queriesadx-mcp-server
- MediumOpen WebUI vulnerable to Path Traversal in `POST /api/v1/audio/transcriptions`open-webui
- Mediumn8n: Authenticated XSS and Open Redirect via Form Noden8n, n8n, n8n
- LowOpen WebUI's Insecure Direct Object Reference (IDOR) allows access to other users' memoriesopen-webui
- MediumOpenClaw: BlueBubbles Webhook Missing Rate Limiting Enables Brute-Force Password Guessingopenclaw
- HighOpenClaw has Inconsistent Host Exec Environment Override Sanitizationopenclaw
- HighOpenClaw's mutating internal ACP chat commands missed operator.admin scope enforcementopenclaw
- HighOpenClaw's Trusted-proxy Control UI sessions retain privileged scopes without device identity on device-less allow pathsopenclaw
- HighOpenClaw: Remote media error responses could trigger unbounded memory allocation before failureopenclaw
- HighOpenClaw: Nostr inbound DMs could trigger unauthenticated crypto work before sender policy enforcementopenclaw
- MediumOpenClaw: Gateway Canvas local-direct requests bypass Canvas HTTP and WebSocket authenticationopenclaw
- MediumOpenClaw's Conflicting Tool Identity Hints Bypass Dangerous-Tool Promptingopenclaw
- MediumOpenClaw: Symlink Traversal via IDENTITY.md appendFile in agents.create/update (Incomplete Fix for CVE-2026-32013)openclaw
- MediumOpenClaw: Forwarding header spoofing bypasses gateway.trustedProxies origin detectionopenclaw
- MediumOpenClaw: Mattermost callback dispatch allowed non-allowlisted sender actionsopenclaw
- Highn8n has SQL Injection in Data Table Node via orderByColumn Expressionn8n, n8n, n8n
- MediumOpenClaw: Image Tool `tools.fs.workspaceOnly` Bypass via Sandbox Bridge Mountsopenclaw
- HighOpenClaw: Plivo V2 verified replay identity drifts on query-only variantsopenclaw
- HighOpenClaw: Arbitrary code execution via unvalidated WebView JavascriptInterfaceopenclaw
- MediumOpenClaw: Windows media loaders accepted remote-host file URLs before local path validationopenclaw
- CriticalOpenClaw Gateway: RCE and Privilege Escalation from operator.pairing to operator.admin via device.pair.approveopenclaw
- HighOpenClaw is vulnerable to Path Traversal through path validation bypassopenclaw
- HighBentoML has Dockerfile Command Injection via system_packages in bentofile.yamlbentoml
- MediumOpenClaw: Google Chat app-url webhook auth accepted non-deployment add-on principalsopenclaw
- Criticaln8n: Prototype Pollution in XML and GSuiteAdmin node parameters lead to RCEn8n, n8n, n8n
- MediumOpenClaw Exposes Credentials Embedded in baseUrl Fields via config.get and channels.statusopenclaw
- LowOpenClaw: Tlon settings empty-allowlist reconciliation bypassed intended revocationopenclaw
- Mediumn8n Vulnerable to XSS via Binary Data Inline HTML Renderingn8n, n8n, n8n
- HighOpenClaw's system.run allowlist can be bypassed through an unregistered time dispatch wrapperopenclaw
- MediumOpenClaw is vulnerable to unauthenticated resource exhaustion through its voice call webhook handlingopenclaw
- MediumOpenClaw Bypasses DM Policy Separation via Synology Chat Webhook Path Collisionopenclaw
- MediumOpenClaw: Bonjour/DNS-SD TXT metadata steers CLI routing after failed service resolutionopenclaw
- CriticalLangflow has Authenticated Code Execution in Agentic Assistant Validationlangflow
- MediumOpenClaw: Tlon cite expansion happens before channel and DM authorization is completeopenclaw
- Mediumn8n Vulnerable to LDAP Filter Injection in LDAP Noden8n, n8n, n8n
- LowOpenClaw may have stale policy enforcement for queued node actionsopenclaw
- HighOpenClaw: Gateway agent /reset exposes admin session reset to operator.write callersopenclaw
- MediumOpenClaw: Synology Chat reply delivery could be rebound through username-based user resolution.openclaw
- MediumOpenClaw leaf subagents can bypass controlScope restrictions to send messages to child sessionsopenclaw
- LowOpenClaw: Nextcloud Talk room allowlist matched colliding room names instead of stable room tokensopenclaw
- Mediumn8n's Source Control SSH Configuration Uses StrictHostKeyChecking=non8n
- Criticaln8n has Multiple Remote Code Execution Vulnerabilities in Merge Node AlaSQL SQL Moden8n, n8n, n8n
- CriticalTwo LiteLLM versions published containing credential harvesting malwarelitellm
- High@grackle-ai/mcp has a workspace authorization bypass in its knowledge_search MCP tool@grackle-ai/mcp
- MediumUnauthenticated SSRF Vulnerability in Streamlit on Windows (NTLM Credential Exposure)streamlit
- Highn8n: LDAP Email-Based Account Linking Allows Privilege Escalation and Account Takeovern8n, n8n
- Highn8n Has External Secrets Authorization Bypass in Credential Savingn8n, n8n
- Highn8n is Vulnerable to Credential Theft via Name-Based Resolution and Permission Checker Bypass in Community Editionn8n, n8n, n8n
- Mediumn8n Has Authorization Bypass in OAuth Callback via N8N_SKIP_AUTH_ON_OAUTH_CALLBACKn8n
- Highn8n has In-Process Memory Disclosure in its Task Runnern8n, n8n, n8n
- Highlangflow has Unauthenticated IDOR on Image Downloadslangflow
- Highlangflow: /profile_pictures/{folder_name}/{file_name} endpoint file readinglangflow
- HighCross-Site Tool Execution for HTTP Servers without Authorizatrion in github.com/modelcontextprotocol/go-sdkgithub.com/modelcontextprotocol/go-sdk
- HighArbitrary file write via tar traversal in mlflowmlflow
- CriticalLangflow has an Arbitrary File Write (RCE) via v2 APIlangflow
- HighClaude Code has a Workspace Trust Dialog Bypass via Repo-Controlled Settings File@anthropic-ai/claude-code
- HighImproper handling of null Unicode character when parsing JSON in github.com/modelcontextprotocol/go-sdkgithub.com/modelcontextprotocol/go-sdk
- CriticalMalware in @iflow-mcp/watercrawl-watercrawl-mcp@iflow-mcp/watercrawl-watercrawl-mcp, @iflow-mcp/watercrawl-watercrawl-mcp, @iflow-mcp/watercrawl-watercrawl-mcp, @iflow-mcp/watercrawl-watercrawl-mcp, @iflow-mcp/watercrawl-watercrawl-mcp
- CriticalMCP Connect has unauthenticated remote OS command execution via /bridge endpointmcp-bridge
- MediumSSRF in @aborruso/ckan-mcp-server via base_url allows access to internal networks@aborruso/ckan-mcp-server
- HighLangflow is Missing Ownership Verification in API Key Deletion (IDOR)langflow
- HighJSONPath Injection in Spring AI Vector Stores FilterExpressionConverterorg.springframework.ai:spring-ai-vector-store, org.springframework.ai:spring-ai-vector-store
- MediumAWS API MCP File Access Restriction Bypassawslabs-aws-api-mcp-server
- HighRay Dashboard is vulnerable to path traversal through its static file handling mechanismray
- CriticalUnauthenticated Remote Code Execution in Langflow via Public Flow Build Endpointlangflow
- HighFastMCP OAuth Proxy token reuse across MCP serversfastmcp
- HighOpenClaw bootstrap setup codes could be replayed to escalate pending pairing scopes before approvalopenclaw
- MediumVanna has a SQL injection in the remove_training_data functionvanna
- HighOpneClaw accepts unsanitized iMessage attachment paths which allowed SCP remote-path command injectionopenclaw
- HighONNX Untrusted Model Repository Warnings Suppressed by silent=True in onnx.hub.load() — Silent Supply-Chain Attackonnx
- HighOpenClaw Telegram webhook request bodies were read before secret validation, enabling unauthenticated resource exhaustionopenclaw
- MediumOpenClaw session transcript files were created without forced user-only permissionsopenclaw
- HighMLflow has a command injection in mlflow/sagemaker/__init__.pymlflow
- MediumOpenClaw Telegram media fetch errors exposed bot tokens in logged file URLsopenclaw
- HighOpenClaw: Gateway `agent` calls could override the workspace boundaryopenclaw
- CriticalOpenClaw: Pairing-scoped device tokens could mint `operator.admin` and reach node RCEopenclaw
- HighOpenClaw: Leaf subagents could steer sibling sessions across sandbox boundariesopenclaw
- Mediumfickling's `platform` module subprocess invocation evades `check_safety()` with `LIKELY_SAFE`fickling
- MediumOpenClaw: Zalo webhook rate limiting could be bypassed before secret validationopenclaw
- MediumOpenClaw: Pairing setup codes exposed long-lived shared gateway credentials instead of short-lived bootstrap tokensopenclaw
- MediumOpenClaw: Channel commands could bypass account-scoped `configWrites` restrictionsopenclaw
- HighOpenClaw: Workspace plugin auto-discovery allowed code execution from cloned repositoriesopenclaw
- MediumOpenClaw: Discord guild reaction ingress could bypass users and roles allowlistsopenclaw
- MediumOpenClaw's Zalouser allowlist authorization matched mutable group names by defaultopenclaw
- MediumOpenClaw: Exec approval allowlist patterns overmatched on POSIX pathsopenclaw
- HighOpenClaw: Feishu webhook mode accepted forged events when only `verificationToken` was configuredopenclaw
- MediumOpenClaw: Write-scoped callers could reach admin-only session reset logic through `agent`openclaw
- MediumOpenClaw: Feishu reaction events could bypass group authorization and mention gatingopenclaw
- HighOpenClaw: Sandbox staged writes could escape the verified parent directory before commitopenclaw
- HighOpenClaw: Unrecognized script runners could bypass `system.run` approval integrityopenclaw
- LowOpenClaw: Unavailable local auth SecretRefs could fall through to remote credentials in local modeopenclaw
- Mediumfickling modules linecache, difflib and gc are missing from the unsafe modules blocklistfickling
- HighOpenClaw: Command-authorized non-owners could reach owner-only `/config` and `/debug` surfacesopenclaw
- CriticalOpenClaw: WebSocket shared-auth connections could self-declare elevated scopesopenclaw
- HighOpenClaw: Node-host approvals could show misleading shell payloads instead of the executed argvopenclaw
- HighOpenClaw: `browser.request` let `operator.write` persist admin-only browser profile changesopenclaw
- High`OpenClaw: session_status` let sandboxed subagents access parent or sibling session stateopenclaw
- HighOpenClaw: Unbound interpreter and runtime commands could bypass node-host approval integrityopenclaw
- MediumOpenClaw: Sandbox `writeFile` commit could race outside the validated pathopenclaw
- CriticalOpenClaw: Plugin subagent routes could bypass gateway authorization with synthetic admin scopesopenclaw
- HighOpenClaw: Untrusted web origins can obtain authenticated operator.admin access in trusted-proxy modeopenclaw
- MediumOpenClaw's system.run approvals did not bind mutable script operands across approval and executionopenclaw
- MediumOpenClaw: /api/channels gateway-auth boundary bypass via path canonicalization mismatchopenclaw
- Mediumha-mcp OAuth 2.1 DCR mode enables network reconnaissance via an error oracleha-mcp
- MediumOpenClaw's MS Teams sender allowlist bypass when route allowlist is configured and sender allowlist is emptyopenclaw
- HighOpenClaw: LINE group allowlist scope mismatch with DM pairing-store entriesopenclaw
- HighSGLangs `replay_request_dump.py` contains an insecure pickle.load() without validation and proper deserializationsglang
- CriticalSGLang's encoder parallel disaggregation system is vulnerable to unauthenticated remote code execution through the disaggregation modulesglang
- HighOpenClaw: workspace path guard bypass on non-existent out-of-root symlink leafopenclaw
- Mediumha-mcp has XSS via Unescaped HTML in OAuth Consent Formha-mcp
- HighOpenClaw: Sandbox dangling-symlink alias handling could bypass workspace-only write boundaryopenclaw
- CriticalSGLang's multimodal generation module is vulnerable to unauthenticated remote code execution through the ZMQ brokersglang
- MediumOpenClaw: Slack system events bypass sender authorization in member and message subtype handlersopenclaw
- MediumOpenClaw's skills-install-download can be redirected outside the tools root by rebinding the validated base pathopenclaw
- HighFlowise affected by Server-Side Request Forgery (SSRF) in HTTP Node Leading to Internal Network Accessflowise, flowise-components
- HighMCP Atlassian has SSRF via unvalidated X-Atlassian-Jira-Url / X-Atlassian-Confluence-Url headersmcp-atlassian
- HighAzure MCP Server has Server-Side Request Forgery issue that allows authorized attacker to elevate privileges over a network@azure/mcp, @azure/mcp, Azure.Mcp, Azure.Mcp
- CriticalMCP Atlassian has an arbitrary file write leading to arbitrary code execution via unconstrained download_path in confluence_download_attachmmcp-atlassian
- MediumOpenClaw's system.run allowlist approval parsing missed PowerShell encoded-command wrappersopenclaw
- HighOpenClaw: fetch-guard forwards custom authorization headers across cross-origin redirectsopenclaw
- MediumOpenClaw's hooks count non-POST requests toward auth lockoutopenclaw
- MediumOpenClaw: system.run allow-always persistence included shell-commented payload tailsopenclaw
- MediumOpenClaw: Sandboxed /acp spawn requests could initialize host ACP sessionsopenclaw
- MediumOpenClaw: `operator.write` chat.send could reach admin-only config writesopenclaw
- MediumOpenClaw's `system.run` env override filtering allowed dangerous helper-command pivotsopenclaw
- MediumOpenClaw: Cross-account sender authorization expansion in `/allowlist ... --store` account scopingopenclaw
- LowOpenClaw: system.run wrapper-depth boundary could skip shell approval gatingopenclaw
- HighOpenClaw's dashboard leaked gateway auth material via browser URL/query and localStorageopenclaw
- MediumvLLM has SSRF Protection Bypassvllm
- Highmcp-memory-service's Wildcard CORS with Credentials Enables Cross-Origin Memory Theftmcp-memory-service
- HighFlowise Missing Authentication on NVIDIA NIM Endpointsflowise
- HighFlowise has IDOR leading to Account Takeover and Enterprise Feature Bypass via SSO Configurationflowise
- HighGitHub Copilot CLI Dangerous Shell Expansion Patterns Enable Arbitrary Code Execution@github/copilot
- HighFlowise has Arbitrary File Upload via MIME Spoofingflowise
- HighFlowise Allows Mass Assignment in `/api/v1/leads` Endpointflowise
- HighFlowise has Authorization Bypass via Spoofed x-request-from Headerflowise
- Mediummcp-memory-service Vulnerable to System Information Disclosure via Health Endpointmcp-memory-service
- MediumLangGraph checkpoint loading has unsafe msgpack deserializationlanggraph
- MediumFlowise Vulnerable to PII Disclosure on Unauthenticated Forgot Password Endpointflowise
- MediumFlowise has Insufficient Password Salt Roundsflowise
- HighOpenClaw: Hardlink alias checks could bypass workspace-only file boundaries in specific configurationsopenclaw
- MediumOpenClaw has incomplete IPv4 special-use SSRF blocking in web fetch guardopenclaw
- HighFickling missing RCE-capable modules in UNSAFE_IMPORTSfickling
- MediumOpenClaw has SSRF guard bypass via IPv6 transition over ISATAPopenclaw
- MediumOpenClaw has agent avatar symlink traversal in gateway session metadataopenclaw
- MediumOpenClaw's elevated allowFrom accepted broader identity signals than specified within sender-scoped authorizationopenclaw
- MediumOpenClaw Vulnerable to Local File Exfiltration via MCP Tool Result MEDIA: Directive Injectionopenclaw
- MediumOpenClaw: BlueBubbles (optional plugin) pairing/allowlist mismatch when allowFrom is emptyopenclaw
- MediumOpenClaw's image tool bypasses tools.fs.workspaceOnly on sandbox mount paths and exfiltrates out-of-workspace imagesopenclaw
- LowOpenClaw has cross-account DM pairing authorization bypass via unscoped pairing store accessopenclaw
- HighOpenClaw Canvas Authentication Bypass Vulnerabilityopenclaw
- HighFickling has `always_check_safety()` bypass: pickle.loads and _pickle.loads remain unhookedfickling
- HighOpenClaw: Slack interactive callbacks could skip configured sender checks in some shared-workspace flowsopenclaw
- HighOpenClaw's andbox browser noVNC observer lacked VNC authenticationopenclaw
- MediumOpenClaw has a BlueBubbles group allowlist mismatch via DM pairing-store fallbackopenclaw
- MediumOpenClaw has a workspace-only sandbox guard mismatch for @-prefixed absolute pathsopenclaw
- MediumOpenClaw: Node camera URL payload host-binding bypass allowed gateway fetch pivotsopenclaw
- HighOpenClaw's commands.allowFrom sender authorization accepted conversation identifiers via ctx.Fromopenclaw
- HighOpenClaw has system.run shell-wrapper env injection via SHELLOPTS/PS4 can bypass allowlist intent (RCE)openclaw
- MediumOpenClaw's owner-only gateway tool access checks were incomplete in specific authenticated DM flowsopenclaw
- MediumOpenClaw affected by iMessage remote attachment SCP hardening (strict host-key checks and remoteHost validation)openclaw
- MediumOpenClaw improperly parses X-Forwarded-For behind trusted proxies allows client IP spoofing in security decisionsopenclaw
- LowOpenClaw Vulnerable to HTML injection via unvalidated image MIME type in data-URL interpolationopenclaw
- MediumTemporary path handling could write outside OpenClaw temp boundaryopenclaw
- MediumOpenClaw: Discord DM reaction ingress missed dmPolicy/allowFrom checks in restricted setupsopenclaw
- HighOpenClaw's tools.exec.safeBins sort long-option abbreviation bypass can skip exec approval in allowlist modeopenclaw
- MediumOpenClaw has an opt-in insecure Control UI auth over plaintext HTTP could allow privileged accessopenclaw
- MediumOpenClaw: Unified root-bound write hardening for browser output and related path-boundary flowsopenclaw
- MediumOpenClaw exec allowlist safeBins short-option bypass could permit arbitrary file writeopenclaw
- MediumOpenClaw safeBins grep -e File Read Bypass (stdin-only policy bypass)openclaw
- MediumOpenClaw: Chrome --no-sandbox disabled OS-level browser sandbox in sandbox browser containeropenclaw
- MediumOpenClaw browser navigation guard allowed non-network URL schemes, enabling authenticated browser-tool users to access file:// local filesopenclaw
- HighOpenClaw: Sandboxed sessions_spawn(runtime="acp") bypassed sandbox inheritance and allowed host ACP initializationopenclaw
- MediumOpenClaw has allowlist exec-guard bypass via env -Sopenclaw
- MediumOpenClaw has a Discord `allowFrom` slug-collision authorization bypassopenclaw
- MediumIn OpenClaw, manually adding sort to tools.exec.safeBins could bypass allowlist approval via --compress-programopenclaw
- MediumOpenClaw's Zalo group sender allowlist bypass permits unauthorized GROUP dispatchopenclaw
- MediumOpenClaw unpaired device identity can bypass operator pairing and self-assign operator scopes with shared authopenclaw
- MediumOpenClaw vulnerable to arbitrary file read via $include directiveopenclaw
- MediumOpenClaw has hook auth rate limiter bypass via IPv4-mapped IPv6 client key variantsopenclaw
- HighOpenClaw: safeBins static default trusted dirs allow writable-dir binary hijack (`jq`)openclaw
- MediumOpenClaw shell-env fallback trusted startup env and could execute attacker-influenced login-shell pathsopenclaw
- MediumOpenClaw: BlueBubbles beta plugin webhook auth hardening (remove passwordless fallback)openclaw
- LowOpenClaw's runtime /debug override path accepted prototype-reserved keysopenclaw
- HighOpenClaw hook transform path containment missed symlink-resolved escapesopenclaw
- MediumOpenClaw: macOS beta onboarding exposed PKCE verifier via OAuth stateopenclaw
- MediumOpenClaw's system.run shell-wrapper positional argv carriers could execute hidden commands under misleading approval textopenclaw
- MediumOpenClaw skills-install-download: tar.bz2 extraction bypassed archive safety parity checks (local DoS)openclaw
- MediumOpenClaw's Signal reaction-only status events could, in limited cases, be enqueued before access checksopenclaw
- MediumOpenClaw has exec allowlist/safeBins policy-runtime mismatch via env -S wrapper interpretationopenclaw
- MediumOpenClaw: macOS optional allowlist basename matching could bypass path-based policyopenclaw
- HighOpenClaw has command injection via Windows shell fallback in Lobster tool executionopenclaw
- HighOpenClaw's non-default autoAllowSkills setting could bypass on-miss exec promptopenclaw
- LowOpenClaw: Microsoft Teams media fetch paths bypass shared SSRF guard modelopenclaw
- CriticalPickleScan's profile.run blocklist mismatch allows exec() bypasspicklescan
- HighOpenClaw's hook transform module path allows traversal and arbitrary JavaScript module loadingopenclaw
- MediumOpenClaw's config env vars allowed startup env injection into service runtimeopenclaw
- LowOpenClaw's tools.exec.safeBins generic fallback allowed interpreter-style inline payload execution in allowlist modeopenclaw
- MediumOpenClaw's web tools strict URL guard could lose DNS pinning when env proxy is configuredopenclaw
- MediumOpenClaw's system.run allowlist bypass via shell line-continuation command substitutionopenclaw
- HighOpenClaw: Native prompt image auto-load did not honor tools.fs.workspaceOnly in sandboxed runsopenclaw
- HighOpenClaw has macOS `system.run` allowlist bypass via quoted command substitutionopenclaw
- MediumOpenClaw's dispatch-wrapper depth-cap mismatch can bypass shell-wrapper approval gating in system.run allowlist modeopenclaw
- HighOpenClaw: stageSandboxMedia destination symlink traversal can overwrite files outside sandbox workspaceopenclaw
- MediumOpenClaw has Canvas route hardening for mixed-trust deploymentsopenclaw
- HighOpenClaw's shell env fallback trusts unvalidated SHELL path from host environmentopenclaw
- MediumOpenClaw plugin runtime command execution is part of trusted plugin boundaryopenclaw
- MediumOpenClaw has Windows Lobster shell fallback command injection in constrained fallback pathopenclaw
- CriticalPickleScan has multiple stdlib modules with direct RCE not in blocklistpicklescan
- HighOpenClaw's `tools.exec.safeBins` PATH-hijack allowed trojan binaries to bypass allowlist checksopenclaw
- LowOpenClaw's voice-call Twilio replay dedupe now bound to authenticated webhook identityopenclaw
- MediumOpenClaw's serialize sandbox registry writes to prevent races and delete-rollback corruptionopenclaw
- MediumOpenClaw's Synology Chat dmPolicy=allowlist failed open on empty allowedUserIds, allowing unauthorized agent dispatchopenclaw
- MediumOpenClaw's Node system.run approval hardening wrapper semantic drift can execute unintended local scriptsopenclaw
- MediumOpenClaw: Unauthorized Telegram Senders Trigger Media Download and Disk Write Before Access Checkopenclaw
- MediumOpenClaw has a IPv6 multicast SSRF classifier bypassopenclaw
- HighOpenClaw: Experimental apply_patch may bypass workspace-only checks in opt-in sandbox mounts (off by default)openclaw
- MediumOpenClaw's gateway tokenless Tailscale auth applied to HTTP routesopenclaw
- MediumOpenClaw: MS Teams fileConsent/invoke missing conversation binding allowed cross-conversation pending-upload consumptionopenclaw
- MediumOpenClaw has a Feishu allowFrom authorization bypass via display-name collisionopenclaw
- HighOpenClaw's exec allowlist wrapper analysis did not unwrap env/shell dispatch chainsopenclaw
- MediumOpenClaw's browser-origin WebSocket auth hardening gap could enable loopback password brute-force chainsopenclaw
- HighOpenClaw DM pairing-store identities could satisfy group allowlist authorizationopenclaw
- HighOpenClaw: Zip extraction symlink traversal could write outside destinationopenclaw
- HighBentoML Vulnerable to Arbitrary File Write via Symlink Path Traversal in Tar Extractionbentoml
- HighOpenClaw's sandbox bind validation could bypass allowed-root and blocked-path checks via symlink-parent missing-leaf pathsopenclaw
- HighOpenClaw Windows Scheduled Task script generation allowed local command injection via unsafe cmd argument handlingopenclaw
- MediumOpenClaw's system.run approval TOCTOU via mutable symlink cwd target on node hostopenclaw
- HighOpenClaw has gateway plugin auth bypass via encoded dot-segment traversal in protected /api/channels pathsopenclaw
- MediumOpenClaw: shell-env trusted-prefix fallback allowed attacker-controlled binary execution via $SHELLopenclaw
- MediumOpenClaw has auth inconsistency on local Browser Extension Relay /extension endpointopenclaw
- HighOpenClaw has a Command Injection via unescaped environment assignments in Windows Scheduled Task script generationopenclaw
- MediumOpenClaw's tools.exec.safeBins trusted PATH directories allowed binary shadowing in allowlist modeopenclaw
- HighOpenClaw's Telegram message_reaction authorization bypass allows unauthorized system-event injectionopenclaw
- MediumOpenClaw has stored XSS in exported session HTML viewer via markdown/raw-HTML renderingopenclaw
- HighOpenClaw: ZIP extraction race could write outside destination via parent symlink rebindopenclaw
- HighOpenClaw: Node reconnect metadata spoofing could bypass platform-based node command policyopenclaw
- MediumOpenClaw's Nextcloud Talk webhook replay could trigger duplicate inbound processingopenclaw
- MediumOpenClaw's Slack reaction/pin sender-policy consistency issue in non-message ingressopenclaw
- MediumOpenClaw's Node role device-identity bypass allows unauthorized node.event injectionopenclaw
- MediumOpenClaw's avatar symlink traversal can expose out-of-workspace local filesopenclaw
- MediumOpenClaw Loopback CDP probe can leak Gateway token to local listeneropenclaw
- LowOpenClaw reuses the gateway auth token in the owner ID prompt hashing fallbackopenclaw
- HighOpenClaw has encoded-path auth bypass in plugin `/api/channels` route classificationopenclaw
- HighOpenClaw Improperly Neutralizes Line Breaks in systemd Unit Generation Enables Local Command Execution (Linux)openclaw
- MediumOpenClaw vulnerable to path traversal in Feishu media temp-file naming allows writes outside os.tmpdir()openclaw
- MediumOpenClaw's non-default safeBins sort configuration can bypass intended allowlist approval constraintsopenclaw
- MediumOpenClaw's voice-call Twilio webhook replay could bypass manager dedupe because normalized event IDs were randomized per parseopenclaw
- MediumOpenClaw has a Trusted-proxy Control UI pairing bypass which allows unpaired node sessionsopenclaw
- CriticalPickleScan's pkgutil.resolve_name has a universal blocklist bypasspicklescan
- HighOpenClaw's MSTeams attachment redirect handling could bypass configured media host allowlistsopenclaw
- HighOpenClaw has multiple E2E/test Dockerfiles that run all processes as rootopenclaw
- HighOpenClaw affected by BASH_ENV / ENV startup-file injection into spawned shell commandsopenclaw
- MediumOpenClaw: Discord voice transcript owner-flag omission could expose owner-only tools in mixed-trust channelsopenclaw
- MediumOpenClaw's typed sender-key matching for toolsBySender prevents identity-collision policy bypassopenclaw
- MediumOpenClaw has pre-auth webhook body parsing that can enable unauthenticated slow-request DoSopenclaw
- HighOpenClaw vulnerable to sensitive file disclosure via stageSandboxMediaopenclaw
- MediumMCP NMAP Server has an Injection vulnerabilitymcp-nmap-server
- HighOpenClaw's shell startup env injection bypasses system.run allowlist intent (RCE class)openclaw
- HighOpenClaw: Sandbox media fallback tmp symlink alias bypass allows host file reads outside sandboxRootopenclaw
- MediumOpenClaw has browser trace/download path symlink escape in temp output handlingopenclaw
- MediumOpenClaw: Unicode canonicalization drift in node metadata policy classification could broaden node allowlistsopenclaw
- MediumOpenClaw: Config writes could persist resolved ${VAR} secrets to diskopenclaw
- LowOpenClaw macOS companion app (beta): allowlist parsing mismatch for system.run shell chainsopenclaw
- LowOpenClaw's Control UI Static File Handler Follows Symlinks and Allows Out-of-Root File Readopenclaw
- HighOpenClaw has Windows system.run approval mismatch on cmd.exe /c trailing argumentsopenclaw
- CriticalCpenClaw's ACPX Windows wrapper shell fallback allowed cwd injection in specific pathsopenclaw
- MediumOpenClaw's allow-always wrapper persistence could bypass future approvals and enable command executionopenclaw
- MediumOpenClaw: Node exec approvals could be replayed across nodesopenclaw
- HighOpenClaw: Sandbox media TOCTOU could read files outside sandbox rootopenclaw
- MediumOpenClaw has an unauthorized sender bypass in its stop triggers and /models command authorizationopenclaw
- HighOpenClaw: Gateway /tools/invoke tool escalation + ACP permission auto-approvalopenclaw
- MediumOpenClaw has ReDoS and regex injection via unescaped Feishu mention metadata in RegExp constructionopenclaw
- HighOpenClaw: Node system.run approval bypass via parent-symlink cwd rebindopenclaw
- CriticalOpenClaw gateway agents.files symlink escape allowed out-of-workspace file read/writeopenclaw
- HighOpenClaw: Message action attachment hydration bypasses local media root checks when sandboxRoot is unsetopenclaw
- HighOpenClaw has web_search citation redirect SSRF via private-network-allowing policyopenclaw
- CriticalOpenClaw Vulnerable to Remote Code Execution via Node Invoke Approval Bypass in Gatewayopenclaw
- MediumOpenClaw's exec allow-always can be bypassed via unrecognized multiplexer shell wrappers (busybox/toybox sh -c)openclaw
- LowOpenClaw Node system.run approval context-binding weakness in approval-enabled host=node flowsopenclaw
- MediumOpenClaw: system.run approval identity mismatch could execute a different binary than displayedopenclaw
- HighOpenClaw has non-constant-time token comparison in hooks authenticationopenclaw
- HighOpenClaw Canvas Path Traversal Information Disclosure Vulnerabilityopenclaw
- HighOpenClaw's authorization mismatch allowed write-scope agent runs to reach owner-only toolsopenclaw
- HighOpenClaw voice-call media stream validated streams after upgrade, which could allow pre-start unauthenticated sockets to increase resource p@openclaw/voice-call, openclaw
- HighOpenClaw has Zip Slip path traversal in tar archive extractionopenclaw
- MediumOpenClaw's sandboxed sessions_spawn now enforces sandbox inheritance for cross-agent spawnsopenclaw
- HighOpenClaw: system.run approvals did not bind PATH-token executable identity, enabling post-approval executable rebindopenclaw
- HighOpenClaw's inbound media downloads could exceed configured byte limits before rejection across multiple channelsopenclaw
- MediumOpenClaw vulnerable to path traversal (Zip Slip) in archive extraction during explicit installation commandsopenclaw
- HighOpenClaw: Browser control startup could continue unauthenticated after auth bootstrap failureopenclaw
- LowOpenClaw has Signal group allowlist authorization bypass via DM pairing-store leakageopenclaw
- MediumOpenClaw has unbounded memory growth in Zalo webhook via query-string key churn (unauthenticated DoS)openclaw
- MediumOpenClaw has a sandbox network isolation bypass via docker.network=container:<id>openclaw
- HighOpenClaw's TOCTOU symlink race in writeFileWithinRoot could create or truncate files outside root boundariesopenclaw
- MediumOpenClaw's sandbox skill mirroring path traversal vulnerability could write outside the sandbox workspaceopenclaw
- HighGradio is Vulnerable to Absolute Path Traversal on Windows with Python 3.13+gradio
- LowGradio: Mocked OAuth Login Exposes Server Credentials and Uses Hardcoded Session Secretgradio
- HighGradio has SSRF via Malicious `proxy_url` Injection in `gr.load()` Config Processinggradio
- MediumGradio has an Open Redirect in its OAuth Flowgradio
Feb 2026163 advisories
- CriticalLangflow has Remote Code Execution in CSV Agentlangflow
- CriticalOpenClaw is vulnerable to validation bypass through GNU long-option abbreviations in allowlist modeopenclaw
- MediumOpenClaw ACP client has permission auto-approval bypass via untrusted tool metadataopenclaw
- MediumOpenClaw: Skill env override host env injection via applySkillConfigEnvOverrides (defense-in-depth)openclaw
- Mediumn8n has Webhook Forgery on Zendesk Trigger Noden8n, n8n
- Mediumn8n: SQL Injection in MySQL, PostgreSQL, and Microsoft SQL nodesn8n
- Mediumn8n has a Guardrail Node Bypassn8n
- Mediumn8n has an Authentication Bypass in its Chat Trigger Noden8n, n8n, n8n
- Mediumn8n: Webhook Forgery on Github Webhook Triggern8n, n8n
- Mediumn8n has an SSO Enforcement Bypass in its Self-Service Settings APIn8n
- Mediummcp-server-git : Path traversal in git_add allows staging files outside repository boundariesmcp-server-git
- HighMCP Go SDK Vulnerable to Improper Handling of Case Sensitivitygithub.com/modelcontextprotocol/go-sdk
- Highn8n Vulnerable to Stored XSS via Various Nodesn8n, n8n, n8n
- Criticaln8n has Unauthenticated Expression Evaluation via Form Noden8n, n8n, n8n
- Criticaln8n has a Sandbox Escape in its JavaScript Task Runnern8n, n8n, n8n
- MediumFickling has safety check bypass via REDUCE+BUILD opcode sequencefickling
- MediumLangGraph: BaseCache Deserialization of Untrusted Data may lead to Remote Code Executionlanggraph-checkpoint
- Highn8n has Arbitrary File Read via Python Code Node Sandbox Escapen8n, n8n, n8n
- MediumLangChain Community: redirect chaining can lead to SSRF bypass via RecursiveUrlLoader@langchain/community
- Criticaln8n: Expression Sandbox Escape Leads to RCEn8n, n8n, n8n
- Criticaln8n has Potential Remote Code Execution via Merge Noden8n, n8n, n8n
- Criticaln8n has Arbitrary Command Execution via File Write and Git Operationsn8n, n8n
- HighFickling: OBJ opcode call invisibility bypasses all safety checksfickling
- MediumOC-91: WhatsApp reaction action bypasses JID authorization for allowFrom-restricted targetsopenclaw
- HighShell command injection in exec allowlist mode via unquoted heredoc expansionopenclaw
- MediumACP resource_link metadata prompt interpolation allowed prompt-injection content via title/URI fieldsopenclaw
- CriticalMLflow Use of Default Password Authentication Bypass Vulnerabilitymlflow
- HighMLflow Tracking Server Artifact Handler Directory Traversal Remote Code Execution Vulnerabilitymlflow
- LowOC-65: overflow compaction retry counter reset after truncation amplifies bounded retry/cost cyclesopenclaw
- MediumTTS model directives allowed provider switching by defaultopenclaw
- LowFickling has a detection bypass via stdlib network-protocol constructorsfickling
- MediumOpenClaw: ACP prompt-size checks missing in local stdio bridge could reduce responsiveness with very large inputsopenclaw
- MediumRay dashboard DELETE endpoints allow unauthenticated browser-triggered DoS (Serve shutdown / job deletion)ray
- MediumOpenClaw: Reject symlinks in local skill packaging scriptopenclaw
- MediumOpenClaw hardened cron webhook delivery against SSRFopenclaw
- LowOpenClaw Discord moderation authorization used untrusted sender identity in tool-driven flowsopenclaw
- LowOpenClaw safeBins stdin-only bypass via sort output and recursive grep flagsopenclaw
- MediumOpenClaw safeBins file-existence oracle information disclosureopenclaw
- HigheBay API MCP Server Affected by Environment Variable Injectionebay-mcp
- HighOpenClaw replaced a deprecated sandbox hash algorithmopenclaw
- MediumOpenClaw has a Web Fetch DoS via unbounded response parsingopenclaw
- HighOpenClaw has a path traversal in apply_patch could write/delete files outside the workspaceopenclaw
- CriticalMicrosoft Semantic Kernel InMemoryVectorStore filter functionality vulnerable to remote code executionsemantic-kernel
- HighOpenClaw: Unsanitized CWD path injection into LLM promptsopenclaw
- MediumOpenClaw affected by Stored XSS in Control UI via unsanitized assistant name/avatar in inline script injectionopenclaw
- HighOpenClaw affected by cross-site request forgery (CSRF) through loopback browser mutation endpointsclawdbot, openclaw
- HighKeras has a Local File Disclosure via HDF5 External Storage During Keras Weight Loadingkeras, keras
- HighOpenClaw: Prevent shell injection in macOS keychain credential writeopenclaw
- HighOpenClaw's unsanitized session ID enables path traversal in transcript file operationsopenclaw
- MediumOpenClaw session tool visibility hardening and Telegram webhook secret fallbackopenclaw
- MediumOpenClaw Chutes manual OAuth state validation bypass can cause credential substitutionopenclaw
- HighPicklescan (scan_pytorch) Bypass via dynamic eval MAGIC_NUMBERpicklescan
- MediumOpenClaw Twilio voice-call webhook auth bypass when ngrok loopback compatibility is enabledopenclaw
- MediumOpenClaw: Telegram bot token exposure via logsopenclaw
- HighOpenClaw has a path traversal in browser upload allows local file readopenclaw
- MediumOpenClaw iMessage group allowlist authorization inherited DM pairing-store identitiesclawdbot, openclaw
- HighOpenClaw has a path traversal in browser trace/download output paths may allow arbitrary file writesopenclaw
- MediumOpenClaw hardened the skill download target directory validationopenclaw
- MediumOpenClaw affected by denial of service through unguarded archive extraction allowing high expansion/resource abuse (ZIP/TAR)clawdbot, openclaw
- HighOpenClaw has an authentication bypass in sandbox browser bridge serveropenclaw
- HighOpenClaw affected by denial of service via unbounded URL-backed media fetchopenclaw
- MediumOpenClaw: Process Safety - Unvalidated PID Kill via SIGKILL in Process Cleanupopenclaw
- HighOpenClaw: Command hijacking via unsafe PATH handling (bootstrapping + node-host PATH overrides)openclaw
- LowHugging Face Smolagents has a Server-Side Request Forgery issuesmolagents
- HighOpenClaw has a command injection in maintainer clawtributors updateropenclaw
- MediumOpenClaw Telegram allowlist authorization accepted mutable usernamesclawdbot, openclaw
- MediumOpenClaw affected by SSRF in optional Tlon (Urbit) extension authenticationopenclaw
- HighOpenClaw allows unauthenticated discovery TXT records to steer routing and TLS pinningopenclaw
- HighOpenClaw affected by denial of service via unbounded webhook request body bufferingclawdbot, openclaw
- HighOpenClaw Google Chat shared-path webhook target ambiguity allowed cross-account policy-context misroutingclawdbot, openclaw
- HighOpenClaw has a LFI in BlueBubbles media path handlingopenclaw
- HighOpenClaw affected by potential code execution via unsafe hook module path handling in Gatewayopenclaw
- HighOpenClaw Slack: dmPolicy=open allowed any DM sender to run privileged slash commandsopenclaw
- HighOpenClaw: Docker container escape via unvalidated bind mount config injectionopenclaw
- MediumOpenClaw: denial of service through large base64 media files allocating large buffers before limit checksclawdbot, openclaw
- HighOpenClaw inter-session prompts could be treated as direct user instructionsopenclaw
- HighOpenClaw has two SSRF via sendMediaFeishu and markdown image fetching in Feishu extensionopenclaw
- HighOpenClaw exec approvals: safeBins could bypass stdin-only constraints via shell expansionopenclaw
- MediumOpenClaw has a Path Traversal in Browser Download Functionalityopenclaw
- MediumOpenClaw's sandbox config hash sorted primitive arrays and suppressed needed container recreationopenclaw
- HighOpenClaw Twitch allowFrom is not enforced in optional plugin, unauthorized chat users can trigger agent pipelineopenclaw
- HighOpenClaw has an exec allowlist bypass via command substitution/backticks inside double quotesopenclaw
- HighOpenClaw optional voice-call plugin: webhook verification may be bypassed behind certain proxy configurations@clawdbot/voice-call, @openclaw/voice-call
- HighOpenClaw is Missing Webhook Authentication in Telnyx Provider Allows Unauthenticated Requestsopenclaw
- CriticalOpenClaw has an inbound allowlist policy bypass in voice-call extension (empty caller ID + suffix matching)openclaw
- HighOpenClaw affected by SSRF in Image Tool Remote Fetchopenclaw
- HighOpenClaw has an arbitrary transcript path file write via gateway sessionFileopenclaw
- LowOpenClaw Affected by Remote Code Execution via System Prompt Injection in Slack Channel Descriptionsopenclaw
- HighOpenClaw macOS deep link confirmation truncation can conceal executed agent messageopenclaw
- MediumOpenClaw MS Teams inbound attachment downloader leaks bearer tokens to allowlisted suffix domainsopenclaw
- HighOpenClaw has a local file disclosure via sendMediaFeishu in Feishu extensionopenclaw
- MediumOpenClaw skills.status could leak secrets to operator.read clientsopenclaw
- LowOpenClaw Google Chat spoofing access with allowlist authorized mutable email principal despite sender-ID mismatchclawdbot, openclaw
- CriticalOpenClaw has a potential access-group authorization bypass if channel type lookup failsopenclaw
- LowOpenClaw log poisoning (indirect prompt injection) via WebSocket headersopenclaw
- HighOpenClaw Gateway tool allowed unrestricted gatewayUrl overrideopenclaw
- HighOpenClaw Node host system.run rawCommand/command mismatch can bypass allowlist/approvalsopenclaw
- HighOpenClaw Hook Session Key Override Enables Targeted Cross-Session Routingopenclaw
- HighOpenClaw has a SSRF guard bypass via full-form IPv4-mapped IPv6 (loopback / metadata reachable)openclaw
- HighOpenClaw has a Telegram webhook request forgery (missing `channels.telegram.webhookSecret`) → auth bypassopenclaw
- HighOpenClaw authorization bypass: operator.write can resolve exec approvals via chat.send -> /approveopenclaw
- HighOpenClaw's Browser Relay /cdp websocket is missing auth which could allow cross-tab cookie accessmoltbot, openclaw
- MediumOpenClaw's unauthenticated Nostr profile HTTP endpoints allow remote profile/config tamperingopenclaw
- HighOpenClaw BlueBubbles webhook auth bypass via loopback proxy trust@openclaw/bluebubbles, openclaw
- HighOpenClaw's Windows cmd.exe parsing may bypass exec allowlist/approval gatingopenclaw
- HighOpenClaw has a Path Traversal in Plugin Installationopenclaw
- MediumOpenClaw's Chrome extension relay binds publicly due to wildcard treated as loopbackopenclaw
- CriticalNextcloud Talk allowlist bypass via actor.name display name spoofing@openclaw/nextcloud-talk
- MediumOpenClaw has a Matrix allowlist bypass via displayName and cross-homeserver localpart matchingopenclaw
- CriticalOpenClaw's gateway connect could skip device identity checks when auth.token was present but not yet validatedopenclaw
- MediumOpenClaw affected by SSRF via attachment/media URL hydrationopenclaw
- HighOpenClaw has a webhook auth bypass when gateway is behind a reverse proxy (loopback remoteAddress trust)openclaw
- HighSandbox escape via Git hookscursor
- MediumCloudflare Agents is Vulnerable to Reflected Cross-Site Scripting in the AI Playground's OAuth callback handleragents
- MediumCloudflare Agents has a Reflected Cross-Site Scripting (XSS) vulnerability in AI Playground siteagents
- LowLangChain affected by SSRF via image_url token counting in ChatOpenAI.get_num_tokens_from_messageslangchain-core
- CriticalMilvus: Unauthenticated Access to Restful API on Metrics Port (9091) Leads to Critical System Compromisegithub.com/milvus-io/milvus, github.com/milvus-io/milvus
- Medium@langchain/community affected by SSRF Bypass in RecursiveUrlLoader via insufficient URL origin validation@langchain/community
- MediumMCP Run Python Deno Sandbox Misconfiguration Allows SSRF Attacks via Localhost Accessmcp-run-python
- MediumMCP Run Python has a Sandbox Escape & Server Takeover Vulnerabilitymcp-run-python
- MediumLangSmith Client SDK Affected by Server-Side Request Forgery via Tracing Header Injectionlangsmith
- Mediummcp-maigret vulnerable to command injectionmcp-maigret
- Lowxcode-mcp-server vulnerable to Command Injectionxcode-mcp-server
- HighPydantic AI has Server-Side Request Forgery (SSRF) in URL Download Handlingpydantic-ai, pydantic-ai-slim
- CriticalSemantic Kernel has Arbitrary File Write via AI Agent Function Calling in .NET SDKMicrosoft.SemanticKernel.Core, semantic-kernel
- LowClaude Code has Permission Deny Bypass Through Symbolic Links@anthropic-ai/claude-code
- HighClaude Code Vulnerable to Command Injection via Directory Change Bypasses Write Protection@anthropic-ai/claude-code
- HighClaude Code has Sandbox Escape via Persistent Configuration Injection in settings.json@anthropic-ai/claude-code
- HighClaude Code Vulnerable to Command Injection via Piped sed Command Bypasses File Write Restrictions@anthropic-ai/claude-code
- HighMCP-Salesforce's arbitrary attribute access leads to disclosure of Salesforce auth tokenmcp-salesforce-connector
- HighPydantic AI has Stored XSS via Path Traversal in Web UI CDN URLpydantic-ai, pydantic-ai-slim
- Highqdrant has arbitrary file write via `/logger` endpointqdrant
- Mediumn8n's domain allowlist bypass enables credential exfiltrationn8n
- High@modelcontextprotocol/sdk has cross-client data leak via shared server/transport instance reuse@modelcontextprotocol/sdk
- Highn8n's Unsafe Buffer Allocation Allows In-Process Memory Disclosure in Task Runnern8n
- Criticaln8n Has Expression Escape Vulnerability Leading to RCEn8n, n8n
- Criticaln8n Vulnerable to Command Injection in Community Package Installationn8n
- Highn8n's Improper CSP Enforcement in Webhook Responses May Allow Stored XSSn8n, n8n
- Criticaln8n has a Python sandbox escapen8n
- Highgodot-mcp has Command Injection via unsanitized projectPath@coding-solo/godot-mcp
- Criticaln8n has OS Command Injection in Git Noden8n, n8n
- HighOpenClaw vulnerable to Unauthenticated Local RCE via WebSocket config.applyopenclaw
- Criticaln8n's Improper File Access Controls Allow Arbitrary File Read by Authenticated Usersn8n, n8n
- Criticaln8n Merge Node has Arbitrary File Write leading to RCEn8n, n8n
- Highn8n Vulnerable to Arbitrary File Write on Remote Systems via SSH Noden8n, n8n
- Highn8n Has Stored Cross-site Scripting via Markdown Rendering in Workflow UIn8n, n8n
- MediumOpenClaw Vulnerable to Local File Inclusion via MEDIA: Path Extractionopenclaw
- HighClaude Code has a Path Restriction Bypass via ZSH Clobber which Allows Arbitrary File Writes@anthropic-ai/claude-code
- HighClaude Code has a Command Injection in find Command Bypasses User Approval Prompt@anthropic-ai/claude-code
- MediumCloudflare Agents SDK has Insecure Direct Object Reference (IDOR) via Header-Based Email Routingagents
- HighClaude Code has a Domain Validation Bypass which Allows Automatic Requests to Attacker-Controlled Domains@anthropic-ai/claude-code
- Mediumllama-index-core vulnerable to Uncontrolled Resource Consumptionllama-index-core
- CriticalvLLM has RCE In Video Processingvllm
- Highmlflow Creates of Temporary File in Directory with Insecure Permissionsmlflow
- Highpicklescan missing detection by simple obfuscation of a `builtins.eval` callpicklescan
- HighOpenClaw/Clawdbot has 1-Click RCE via Authentication Token Exfiltration From gatewayUrlclawdbot
- HighHugging Face Text Generation Inference vulnerable to Uncontrolled Resource Consumptiontext-generation
- Mediumpicklescan vulnerable to arbitrary file create using logging.FileHandlerpicklescan
- HighOpenClaw/Clawdbot Docker Execution has Authenticated Command Injection via PATH Environment Variableclawdbot
- HighOpenClaw/Clawdbot has OS Command Injection via Project Root Path in sshNodeCommandclawdbot
- CriticalH2O has an External Control of File Name or Path vulnerabilityh2o
- CriticalLangroid has WAF Bypass Leading to RCE in TableChatAgentlangroid
- MediumLobeHub Vulnerable to Improper Authorization in Presigned Upload@lobehub/chat
Jan 202633 advisories · 1 exploited
- HighvLLM vulnerable to Server-Side Request Forgery (SSRF) through MediaConnectorvllm
- Criticaln8n Unsafe Workflow Expression Evaluation Allows Remote Code Executionn8n, n8n, n8n
- HighPyTorch Vulnerable to Remote Code Execution via Untrusted Checkpoint Filestorch
- HighBentoML has a Path Traversal via Bentofile Configurationbentoml
- HighLangflow affected by Remote Code Execution via validate_code() exec()langflow
- HighvLLM affected by RCE via auto_map dynamic module loading during model initializationvllm
- MediumClaude Code Leaks Data via Malicious Environment Configuration Before Trust Confirmation@anthropic-ai/claude-code
- HighChainlit contain a server-side request forgery (SSRF) vulnerabilitychainlit
- CriticalLobe Chat affected by Cross-Site Scripting(XSS) that can escalate to Remote Code Execution(RCE)@lobehub/chat
- Highchainlit : Chainlit allows authenticated arbitrary file reads through custom element updateschainlit
- LowLobe Chat has IDOR in Knowledge Base File Removal that Allows Cross User File Deletion@lobehub/chat
- HighTerminal Tool Allowlist Bypass via Environment Variablescursor
- LowChainlit contains an authorization bypass vulnerabilitychainlit
- CriticalMalicious website can execute commands on the local system through XSS in the OpenCode web UIopencode-ai
- MediumvLLM is vulnerable to DoS in Idefics3 vision models via image payload with ambiguous dimensionsvllm
- Criticalorval MCP client is vulnerable to a code injection attack.@orval/mcp
- HighOpenCode's Unauthenticated HTTP Server Allows Arbitrary Command Executionopencode-ai
- Mediumn8n: Webhook Node IP Whitelist Bypass via Partial String Matchingn8n
- HighMLFlow is vulnerable to DNS rebinding attacks due to a lack of Origin header validationmlflow
- HighFickling vulnerable to use of ctypes and pydoc gadget chain to bypass detectionfickling
- HighFickling vulnerable to detection bypass due to "builtins" blindnessfickling
- HighFickling Blocklist Bypass: cProfile.run()fickling
- HighFickling has Static Analysis Bypass via Incomplete Dangerous Module Blocklistfickling
- HighFickling has a bypass via runpy.run_path() and runpy.run_module()fickling
- Highpicklescan has Arbitrary file read using `io.FileIO`picklescan
- HighvLLM introduced enhanced protection for CVE-2025-62164vllm
- HighMicrosoft Playwright MCP Server vulnerable to DNS Rebinding Attack; Allows Attackers Access to All Server Tools@playwright/mcp
- Highfast-filesystem-mcp has a Path Traversal vulnerabilityfast-filesystem-mcp
- Mediumn8n's Missing Stripe-Signature Verification Allows Unauthenticated Forged Webhooksn8n
- Criticaln8n Vulnerable to Unauthenticated File Access via Improper Webhook Request Handlingn8n
- Criticaln8n Vulnerable to RCE via Arbitrary File Writen8n
- HighAnthropic's MCP TypeScript SDK has a ReDoS vulnerability@modelcontextprotocol/sdk
- HighLangflow Missing Authentication on Critical API Endpointslangflow, langflow-base
Dec 202554 advisories · 3 malicious packages · 2 exploited
- HighPicklescan is vulnerable to RCE via missing detection when calling built-in python _operator.attrgetterpicklescan
- MediumPicklescan is vulnerable to RCE through missing detection when calling numpy.f2py.crackfortran._eval_lengthpicklescan
- HighPicklescan is vulnerable to RCE via missing detection when calling built-in python _operator.methodcallerpicklescan
- CriticalMalware in vscode-azure-mcp-servervscode-azure-mcp-server
- HighPicklescan is vulnerable to RCE via missing detection when calling numpy.f2py.crackfortran.getlincoefpicklescan
- HighPicklescan is vulnerable to RCE through missing detection when calling numpy.f2py.crackfortran.myevalpicklescan
- HighPicklescan does not block ctypespicklescan
- HighPicklescan has Incomplete List of Disallowed Inputspicklescan
- MediumPicklescan is vulnerable to RCE via missing detection when calling numpy.f2py.crackfortran.param_evalpicklescan
- HighPicklescan Bypasses Unsafe Globals Check using pty.spawnpicklescan
- HighPicklescan vulnerable to Arbitrary File Writingpicklescan
- HighPicklescan missing detection when calling numpy.f2py.crackfortran.getlincoefpicklescan
- HighPicklescan missing detection when calling pty.spawnpicklescan
- HighPicklescan is vulnerable to RCE through missing detection when calling built-in python operator.methodcallerpicklescan
- Highn8n's Possible Stored XSS in "Respond to Webhook" Node May Execute Outside iframe Sandboxn8n
- Criticaln8n Vulnerable to Arbitrary Command Execution in Pyodide based Python Code Noden8n
- Highlmdeploy vulnerable to Arbitrary Code Execution via Insecure Deserialization in torch.load()lmdeploy
- HighSelf-hosted n8n has Legacy Code node that enables arbitrary file read/writen8n
- HighFastMCP updated to MCP 1.23+ due to CVE-2025-66416fastmcp
- CriticalLangChain serialization injection vulnerability enables secret extraction in dumps/loads APIslangchain-core, langchain-core
- CriticalMalware in excluder-mcpexcluder-mcp
- CriticalMalware in excluder-mcp-packageexcluder-mcp-package
- CriticalHugging Face smolagents: Unsafe deserialization in Remote Python Executor leads to RCEsmolagents
- HighLangChain serialization injection vulnerability enables secret extraction@langchain/core, @langchain/core, langchain, langchain
- Criticaln8n Vulnerable to Remote Code Execution via Expression Injectionn8n, n8n
- HighLangflow vulnerable to Server-Side Request Forgerylangflow
- HighExternal Control of File Name or Path in Langflowlangflow
- CriticalOllama Platform has missing authentication enabling attackers to perform model management operationsgithub.com/ollama/ollama
- Highnbconvert has an uncontrolled search path that leads to unauthorized code execution on Windowsnbconvert
- Mediummcp-server-git's unrestricted git_init tool allows repository creation at arbitrary filesystem locationsmcp-server-git
- Mediummcp-server-git argument injection in git_diff and git_checkout functions allows overwriting local filesmcp-server-git
- Mediummcp-server-git has missing path validation when using --repository flagmcp-server-git
- HighFickling has missing detection for marshal.loads and types.FunctionType in unsafe modules listfickling
- HighFickling has Code Injection vulnerability via pty.spawn()fickling
- HighWeaviate OSS has a Path Traversal Vulnerability via Backup ZipSlipgithub.com/weaviate/weaviate, github.com/weaviate/weaviate, github.com/weaviate/weaviate, github.com/weaviate/weaviate
- HighWeaviate OSS has path traversal vulnerability via the Shard Movement APIgithub.com/weaviate/weaviate, github.com/weaviate/weaviate, github.com/weaviate/weaviate, github.com/weaviate/weaviate
- MediumFetch MCP Server has a Server-Side Request Forgery (SSRF) vulnerabilitymcp-fetch-server
- HighLangGraph's SQLite is vulnerable to SQL injection via metadata filter key in SQLite checkpointer list methodlanggraph-checkpoint-sqlite
- Criticaln8n vulnerable to Remote Code Execution via Git Node Custom Pre-Commit Hookn8n
- CriticalLangflow CORS misconfiguration enables Account Takeover and RCElangflow
- MediumComposioHQ has a directory traversal vulnerabilitycomposio
- HighOpen WebUI Vulnerable to Stored DOM XSS via Note 'Download PDF'open-webui
- HighOpen WebUI vulnerable to Server-Side Request Forgery (SSRF) via Arbitrary URL Processing in /api/v1/retrieval/process/webopen-webui
- Lowopen-webui is Vulnerable to Incorrect Access Controlopen-webui
- HighDocker MCP Plugin and Docker MCP Gateway have DNS Rebinding vulnerability when running in sse or streaming modegithub.com/docker/mcp-gateway
- Mediummcp-server-kubernetes has potential security issue in exec_in_pod toolmcp-server-kubernetes
- HighClaude Code Command Validation Bypass Allows Arbitrary Code Execution@anthropic-ai/claude-code
- CriticalMCP Watch has a Critical Command Injection in cloneRepo allows Remote Code Execution (RCE) via malicious URLmcp-watch
- HighvLLM vulnerable to remote code execution via transformers_utils/get_configvllm
- HighModel Context Protocol (MCP) Python SDK does not enable DNS rebinding protection by defaultmcp
- Mediumarcade-mcp-server Has Default Hardcoded Worker Secret That Allows Full Unauthorized Access to All HTTP MCP Worker Endpointsarcade-mcp-server
- MediumPortkey.ai Gateway: Server-Side Request Forgery (SSRF) in Custom Host@portkey-ai/gateway
- HighKeras Directory Traversal Vulnerabilitykeras
- HighModel Context Protocol (MCP) TypeScript SDK does not enable DNS rebinding protection by default@modelcontextprotocol/sdk
Nov 202530 advisories · 9 malicious packages · 1 exploited
- CriticalRay's New Token Authentication is Disabled By Defaultray
- CriticalRay is vulnerable to Critical RCE via Safari & Firefox Browsers through DNS Rebinding Attackray
- CriticalMalware in mcp-polymarketmcp-polymarket
- CriticalMalware in @browserbasehq/mcp@browserbasehq/mcp
- CriticalMalware in @seezo/sdr-mcp-server@seezo/sdr-mcp-server
- CriticalMalware in web-scraper-mcpweb-scraper-mcp
- CriticalMalware in @relyt/mcp-server-relytone@relyt/mcp-server-relytone
- CriticalMalware in ids-enterprise-mcp-serverids-enterprise-mcp-server
- CriticalMalware in @relyt/claude-context-mcp@relyt/claude-context-mcp
- CriticalMalware in lite-serper-mcp-serverlite-serper-mcp-server, lite-serper-mcp-server
- MediumvLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`vllm
- HighLangChain Vulnerable to Template Injection via Attribute Access in Prompt Templateslangchain-core, langchain-core
- High@anthropic-ai/claude-code has Sed Command Validation Bypass that Allows Arbitrary File Writes@anthropic-ai/claude-code
- HighvLLM deserialization vulnerability leading to DoS and potential RCEvllm
- HighvLLM vulnerable to DoS with incorrect shape of multimodal embedding inputsvllm
- HighClaude Code vulnerable to command execution prior to startup trust dialog@anthropic-ai/claude-code
- HighFlowise has Authentication Bypass Using Unprotected Registration Endpoint (/register)flowise
- HighFlowise Fails to Invalidate Existing Sessions After Password Changesflowise
- CriticalMilvus Proxy has a Critical Authentication Bypass Vulnerabilitygithub.com/milvus-io/milvus, github.com/milvus-io/milvus, github.com/milvus-io/milvus, github.com/milvus-io/milvus
- CriticalMalware in rce-poc-test-honor-mcprce-poc-test-honor-mcp
- HighOpen WebUI Affected by an External Model Server (Direct Connections) Code Injection via SSE Eventsopen-webui, open-webui
- LowVercel’s AI SDK's filetype whitelists can be bypassed when uploading filesai, ai
- HighOpen WebUI vulnerable to Stored DOM XSS via prompts when 'Insert Prompt as Rich Text' is enabled resulting in ATO/RCEopen-webui, open-webui
- MediumApache Doris-MCP-Server: Improper Access Control results in bypassing a "read-only" modedoris-mcp-server
- HighLangGraph Checkpoint affected by RCE in "json" mode of JsonPlusSerializerlanggraph-checkpoint
- MediumDSPy does not properly restrict file readsdspy
- HighSensitive File Protection Bypass - Path Manipulation Using Backslashes on Windowscursor
- HighSpeedbump Modal Bypass in Cursor MCP Server Deep-Linkcursor
- HighSensitive File Modification - NTFS Path Quirkscursor
- HighCursorignore Bypass via New Cursorignore Writecursor
Oct 202539 advisories · 3 malicious packages
- HighAgno session state overwrites between different sessions/usersagno
- CriticalMalware in shiprocket-mcpshiprocket-mcp
- Highn8n Vulnerable to Remote Code Execution via Git Node Pre-Commit Hookn8n
- HighMLflow Tracking Server Model Creation Directory Traversal Remote Code Execution Vulnerabilitymlflow, mlflow
- HighMLflow Weak Password Requirements Authentication Bypass Vulnerabilitymlflow
- HighLangGraph SQLite Checkpoint Filter Key SQL Injection POC for SqliteStorelanggraph-checkpoint-sqlite
- HighFastMCP Auth Integration Allows for Confused Deputy Account Takeoverfastmcp
- MediumKeras is vulnerable to arbitrary local file loading and Server-Side Request Forgerykeras
- MediumJenkins MCP Server Plugin does not perform permission checks in multiple MCP toolsio.jenkins.plugins:mcp-server
- MediumFastMCP vulnerable to reflected XSS in client's callback pagefastmcp
- MediumFastMCP vulnerable to windows command injection in FastMCP Cursor installer via server_namefastmcp
- HighLangGraph's SQLite store implementation has a SQL Injection Vulnerabilitylanggraph-checkpoint-sqlite
- MediumHugging Face Smolagents XPath injection vulnerability in the search_item_ctrl_f functionsmolagents
- CriticalKeras framework vulnerable to deserialization of untrusted datakeras
- LowLobe Chat vulnerable to Server-Side Request Forgery with native web fetch module@lobehub/chat
- HighPlaywright downloads and installs browsers without verifying the authenticity of the SSL certificateplaywright
- HighFlowise: Authenticated Command Execution and Sandbox Bypass via Puppeteer and Playwright Packagesflowise
- Highllama-index has Insecure Temporary Filellama-index
- HighFlowise is vulnerable to arbitrary file exposure through its ReadFileToolflowise, flowise-components
- Highn8n: Execute Command Node Allows Authenticated Users to Run Arbitrary Commands on Hostn8n
- CriticalFlowise is vulnerable to arbitrary file write through its WriteFileToolflowise, flowise, flowise-components
- HighFlowiseAI/Flosise has File Upload vulnerabilityflowise
- HighvLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` classvllm
- MediumvLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Servervllm
- CriticalMalware in mcp-server-fixthismcp-server-fixthis
- HighvLLM is vulnerable to timing attack at bearer authvllm
- CriticalFlowise vulnerable to RCE via Dynamic function constructor injectionflowise
- Mediumclearml is vulnerable to Path Traversal through its `safe_extract` functionclearml
- MediumZenML is vulnerable to Path Traversal through its `PathMaterializer` classzenml
- HighClaude Code can execute commands prior to the startup trust dialog@anthropic-ai/claude-code
- MediumFlowise vulnerable to XSSflowise
- LowClaude Code permission deny bypass through symlink@anthropic-ai/claude-code
- MediumFlowise Stored XSS vulnerability through logs in chatbotflowise
- CriticalFlowise is vulnerable to stored XSS via "View Messages" allows credential theft in FlowiseAI admin panelflowise
- HighCursor IDE - Sensitive File Overwrite Bypasscursor
- HighRCE via .code-workspace files using Prompt Injectioncursor
- CriticalMalware in @lanyer640/mcp-runcommand-server@lanyer640/mcp-runcommand-server
- MediumPotential Information Leakage via Mermaid Diagramcursor
- Mediummarimo vulnerable to proxy abuse of /mpl/{port}/marimo
Sept 202545 advisories · 5 malicious packages
- Highfigma-developer-mcp vulnerable to command injection in get_figma_data toolfigma-developer-mcp
- Highllama-index-core insecurely handles temporary filesllama-index-core
- LowJupyterLab LaTeX typesetter links did not enforce `noopener` attributejupyterlab
- HighClaude Code Vulnerable to Arbitrary Code Execution via Plugin Autoloading with Specific Yarn Versions@anthropic-ai/claude-code
- CriticalCommand Injection in adb-mcp MCP Serveradb-mcp
- MediumMastra Docs MCP Server `@mastra/mcp-docs-server` Leads to Information Exposure@mastra/mcp-docs-server
- Mediumlobe-chat has an Open Redirect@lobehub/chat
- MediumHugging Face Transformers vulnerable to Regular Expression Denial of Service (ReDoS) in the AdamWeightDecay optimizertransformers
- CriticalH2O affected by a deserialization vulnerabilityh2o
- HighKeras is vulnerable to Deserialization of Untrusted Datakeras
- HighThe Keras `Model.load_model` method **silently** ignores `safe_mode=True` and allows arbitrary code execution when a `.h5`/`.hdf5` file is lkeras
- HighCodex has sandbox bypass due to bug in path configuration logic@openai/codex
- MediumLobe Chat Desktop vulnerable to Remote Code Execution via XSS in Chat Messages@lobehub/chat
- Medium@sequa-ai/sequa-mcp has Command Injection vulnerability@sequa-ai/sequa-mcp
- CriticalMalware in mcp-knowledge-basemcp-knowledge-base
- CriticalMalware in mcp-knowledge-graphmcp-knowledge-graph
- CriticalFlowise has Remote Code Execution vulnerabilityflowise
- Criticalmcp-kubernetes-server has an OS Command Injection vulnerabilitymcp-kubernetes-server
- CriticalFlowise has unsandboxed remote code execution via Custom MCPflowise
- CriticalFlowiseAI Pre-Auth Arbitrary Code Executionflowise
- CriticalFlowise has an Arbitrary File Readflowise
- MediumHackMD MCP Server has Server-Side Request Forgery (SSRF) vulnerabilityhackmd-mcp
- Mediummcp-kubernetes-server has a Command Injection vulnerabilitymcp-kubernetes-server
- HighFlowiseAI/Flowise has Server-Side Request Forgery (SSRF) vulnerabilityflowise
- MediumStored XSS in n8n LangChain Chat Trigger Node via initialMessages Parametern8n
- CriticalFlowise has arbitrary file access due to missing chat flow id validationflowise
- MediumHugging Face Transformers library has Regular Expression Denial of Servicetransformers
- MediumHugging Face Transformers is vulnerable to ReDoS through its MarianTokenizertransformers
- CriticalMalware in mcp-server-everythingmcp-server-everything
- CriticalMalware in mcp-chat-clientmcp-chat-client
- CriticalFlowise Cloud and Local Deployments have Unauthenticated Password Reset Token Disclosure that Leads to Account Takeoverflowise
- CriticalMalware in mcp-luminmcp-lumin
- HighNeo4j Cypher MCP server is vulnerable to DNS rebindingmcp-neo4j-cypher
- CriticalPicklescan is Vulnerable to Unsafe Globals Check Bypass through Subclass Importspicklescan
- HighClaude Code vulnerable to arbitrary code execution caused by maliciously configured git email@anthropic-ai/claude-code
- CriticalPicklescan Bypass is Possible via File Extension Mismatchpicklescan
- CriticalPicklescan: ZIP archive scan bypass is possible through non-exhaustive Cyclic Redundancy Checkpicklescan
- HighClaude Code rg vulnerability does not protect against approval prompt bypass@anthropic-ai/claude-code
- MediumSGLang Remote Code Execution Vulnerability via Unsafe Deserialization in update_weights_from_tensorsglang
- Critical@akoskm/create-mcp-server-stdio is vulnerable to MCP Server Command Injection through `exec` API@akoskm/create-mcp-server-stdio
- HighMCP Inspector is Vulnerable to Potential Command Execution via XSS When Connecting to an Untrusted MCP Server@modelcontextprotocol/inspector
- HighN8N's Chat Trigger component is vulnerable to XSS@n8n/n8n-nodes-langchain
- HighLangchain Community Vulnerable to XML External Entity (XXE) Attackslangchain-community
- HighClaude Code Vulnerable to Arbitrary Code Execution Due to Insufficient Startup Warning@anthropic-ai/claude-code
- Highmcp-markdownify-server vulnerable to command injection in pptx-to-markdown toolmcp-markdownify-server
Aug 202551 advisories
- MediumAiondaDotCom mcp-ssh command injection vulnerability in SSH operations@aiondadotcom/mcp-ssh
- HighPicklescan is missing detection when calling built-in python idlelib.pyshell.ModifiedInterpreter.runcodepicklescan
- HighPicklescan has a missing detection when calling built-in python idlelib.debugobj.ObjectTreeItempicklescan
- MediumPicklescan is missing detection when calling built-in python cProfile.runpicklescan
- MediumPicklescan is missing detection when calling pytorch function torch.utils.bottleneck.__main__.run_autograd_profpicklescan
- MediumPicklescan has a missing detection when calling built-in python trace.Trace.runpicklescan
- HighPicklescan has a missing detection when calling built-in python profile.Profile.runctxpicklescan
- HighPicklescan has a missing detection when calling built-in python idlelib.autocomplete.AutoComplete.get_entitypicklescan
- HighLlamaIndex affected by a Denial of Service (DOS) in JSONReaderllama-index-core
- HighPicklescan has a missing detection when calling built-in python idlelib.autocomplete.AutoComplete.fetch_completionspicklescan
- HighPicklescan has a missing detection when calling built-in python idlelib.calltip.Calltippicklescan
- MediumPicklescan is missing detection when calling built-in Python cProfile.runctxpicklescan
- MediumPicklescan has a missing detection when calling built-in python library idlelib.calltip.get_entitypicklescan
- MediumPicklescan has a missing detection when calling built-in python code.InteractiveInterpreterpicklescan
- MediumPicklescan has a missing detection when calling built-in python lib2to3.pgen2.grammar.Grammar.loadspicklescan
- MediumPicklescan is missing detection when calling built-in python doctest.debug_scriptpicklescan
- MediumPicklescan has a missing detection when calling built-in python trace.Trace.runctxpicklescan
- HighPicklescan is missing detection when calling built-in python idlelib.pyshell.ModifiedInterpreter.runcommandpicklescan
- MediumPicklescan is missing detection when calling built-in python idlelib.run.Executive.runcodepicklescan
- MediumPicklescan is missing detection when calling built-in python lib2to3.pgen2.pgen.ParserGenerator.make_labelpicklescan
- MediumPicklescan is missing detection when calling built-in python library asyncio.unix_events._UnixSubprocessTransport._startpicklescan
- MediumPicklescan has a missing detection when calling built-in python profile.Profile.runpicklescan
- HighPicklescan is missing detection when calling built-in python ensurepip._run_pippicklescan
- HighLangflow Vulnerable to Privilege Escalation via CLI Superuser Creation (Post-RCE)langflow, langflow-base
- MediumPicklescan missing detection when calling pytorch function torch.utils.bottleneck.__main__.run_cprofilepicklescan
- MediumPicklescan missing detection when calling pytorch function torch._dynamo.guards.GuardBuilder.getpicklescan
- MediumPicklescan missing detection when calling pytorch function torch.fx.experimental.symbolic_shapes.ShapeEnv.evaluate_guards_expressionpicklescan
- MediumPicklescan missing detection when calling pytorch function torch.utils.collect_env.runpicklescan
- MediumPicklescan missing detection when calling pytorch function torch.utils.data.datapipes.utils.decoder.basichandlerspicklescan
- HighPicklescan missing detection when calling pytorch function torch.jit.unsupported_tensor_ops.execWrapperpicklescan
- HighPicklescan is missing detection when calling pytorch function torch.utils._config_module.load_configpicklescan
- HighvLLM has remote code execution vulnerability in the tool call parser for Qwen3-Codervllm
- Lowwong2 mcp-cli Command Injection Vulnerability@wong2/mcp-cli
- Highvllm API endpoints vulnerable to Denial of Service Attacksvllm
- Mediumn8n symlink traversal vulnerability in "Read/Write File" node allows access to restricted filesn8n
- HighStored XSS in n8n Form Trigger allows Account Takeover via injected iframe and video/sourcen8n
- HighClaude Code's Permissive Default Allowlist Enables Unauthorized File Read and Network Exfiltration in Claude Code@anthropic-ai/claude-code
- CriticalFlowise OS command remote code executionflowise
- HighPicklescan has pickle parsing logic flaw that leads to malicious pickle file bypasspicklescan
- HighKeras vulnerable to CVE-2025-1550 bypass via reuse of internal functionalitykeras
- HighSKOPS Card.get_model happily allows arbitrary code executionskops
- MediumOllama allows deletion of arbitrary filesgithub.com/ollama/ollama
- MediumHugging Face Transformers Regular Expression Denial of Service (ReDoS) vulnerabilitytransformers
- Highmcp-package-docs vulnerable to command injection in several toolsmcp-package-docs
- HighClaude Code Research Preview has a Path Restriction Bypass which could allow unauthorized file access@anthropic-ai/claude-code
- HighClaude Code echo command allowed bypass of user approval prompt for command execution@anthropic-ai/claude-code
- HighArbitrary code execution from Cursor Agent through a prompt injection via MCP Special Filescursor
- HighArbitrary code execution from Cursor Agent through a prompt injection via Editor Special Filescursor
- HighModification of MCP Server Definitions Bypasses Manual Re-approvalcursor
- MediumBypassing allow list to execute an arbitrary commandcursor
- MediumMCP Install Deeplink Did Not Show Arguments on User-Dialogcursor
Jul 202527 advisories · 2 malicious packages
- CriticalBentoML SSRF Vulnerability in File Upload Processingbentoml
- Criticalsmolagents has Sandbox Escape Vulnerability in the local_python_executor.py Modulesmolagents
- HighSkops may allow MethodNode to access unexpected object fields through dot notation, leading to arbitrary code execution at load timeskops
- HighSkops has Inconsistent Trusted Type Validation that Enables Hidden `operator` Methods Executionskops
- CriticalMalware in @jaswanthlanka/mcp-proxy@jaswanthlanka/mcp-proxy
- MediumOllama vulnerable to Cross-Domain Token Exposuregithub.com/ollama/ollama
- High@translated/lara-mcp vulnerable to command injection in import_tmx tool@translated/lara-mcp
- CriticalMalware in glean-mcp-serverglean-mcp-server
- HighGitHub Kanban MCP Server vulnerable to Command Injection@sunwood-ai-labs/github-kanban-mcp-server
- MediumTransformers is vulnerable to ReDoS attack through its DonutProcessor classtransformers
- MediumLlamaIndex vulnerable to data loss through hash collisions in its DocugamiReader classllama-index
- Criticalmcp-remote exposed to OS command injection via untrusted MCP server connectionsmcp-remote
- HighNode.js Sandbox MCP Server vulnerability can lead to Sandbox Escape via Command Injectionnode-code-sandbox-mcp
- HighMCP Server Kubernetes vulnerable to command injection in several toolsmcp-server-kubernetes
- HighLlamaIndex vulnerable to Path Traversal attack through its encode_image functionllama-index-core
- MediumLlamaIndex vulnerable to DoS attack through uncontrolled recursive JSON parsingllama-index-core
- MediumTransformers vulnerable to ReDoS attack through its SETTING_RE variabletransformers
- MediumTransformers vulnerable to ReDoS attack through its get_imports() functiontransformers
- MediumLlamaIndex has Incomplete Documentation of Program Execution related to JsonPickleSerializer componentllama-index-core
- LowTransformers's Improper Input Validation vulnerability can be exploited through username injectiontransformers
- MediumTransformers's ReDoS vulnerability in get_configuration_file can lead to catastrophic backtrackingtransformers
- HighMCP Python SDK vulnerability in the FastMCP Server causes validation error, leading to DoSmcp
- HighMCP Python SDK has Unhandled Exception in Streamable HTTP Transport, Leading to Denial of Servicemcp
- Mediumn8n is vulnerable to Improper Authorization through its `/stop` endpointn8n
- Mediumn8n Vulnerable to Denial of Service via Malformed Binary Data Requestsn8n
- High@modelcontextprotocol/server-filesystem vulnerability allows for path validation bypass via colliding path prefix@modelcontextprotocol/server-filesystem, @modelcontextprotocol/server-filesystem
- High@modelcontextprotocol/server-filesystem allows for path validation bypass via prefix matching and symlink handling@modelcontextprotocol/server-filesystem, @modelcontextprotocol/server-filesystem
Jun 202513 advisories · 2 malicious packages · 1 exploited
- High@cyanheads/git-mcp-server vulnerable to command injection in several tools@cyanheads/git-mcp-server
- Mediumn8n allows open redirects via the /signin endpointn8n
- MediumiOS Simulator MCP Command Injection allowed via exec APIios-simulator-mcp
- CriticalMalware in rush-mcp-example-pluginrush-mcp-example-plugin
- HighClaude Code Improper Authorization via websocket connections from arbitrary origins@anthropic-ai/claude-code
- HighLangChain Community SSRF vulnerability exists in RequestsToolkit componentlangchain-community
- MediumMLFlow SSRF via gateway_proxy_handlermlflow, mlflow
- Mediumpython-a2a has a path traversal in the create_workflow functionpython-a2a
- CriticalLangflow Unauth RCElangflow, langflow-base
- CriticalMCP Inspector proxy server lacks authentication between the Inspector client and proxy@modelcontextprotocol/inspector
- MediumPotential Information Leakage using JSON schema in Cursor Agentcursor
- Criticalllama_index vulnerable to SQL Injectionllama-index
- CriticalMalware in yaml-mcp-wrapperyaml-mcp-wrapper
May 202519 advisories
- MediumMarkdownify MCP Server allows attackers to read arbitrary filesmcp-markdownify-server
- MediumGradio Allows Unauthorized File Copy via Path Manipulationgradio
- MediumMarkdownify MCP Server allows Server-Side Request Forgery (SSRF) via the Markdownify.get() functionmcp-markdownify-server
- LowGradio CORS Origin Validation Bypass Vulnerabilitygradio
- LowPotential Timing Side-Channel Vulnerability in vLLM’s Chunk-Based Prefix Cachingvllm
- MediumvLLM DOS: Remotely kill vllm over http with invalid JSON schemavllm
- MediumvLLM allows clients to crash the openai server with invalid regexvllm
- MediumvLLM has a Weakness in MultiModalHasher Image Hashing Implementationvllm
- MediumvLLM vulnerable to Regular Expression Denial of Servicevllm
- MediumvLLM Tool Schema allows DoS via Malformed pattern and type Fieldsvllm
- MediumvLLM has a Regular Expression Denial of Service (ReDoS, Exponential Complexity) Vulnerability in `pythonic_tool_parser.py`vllm
- HighLangroid has a Code Injection vulnerability in LanceDocChatAgent through vector_storelangroid
- CriticalvLLM Allows Remote Code Execution via PyNcclPipe Communication Servicevllm
- CriticalLangroid has a Code Injection vulnerability in TableChatAgentlangroid
- MediumHugging Face Transformers Regular Expression Denial of Servicetransformers
- HighOllama Server Vulnerable to Denial of Service (DoS) Attackgithub.com/ollama/ollama
- HighLlamaIndex Vulnerable to Denial of Service (DoS)llama-index
- HighRemote Code Execution Vulnerability in vLLM Multi-Node Cluster Configurationvllm
- HighLangroid Allows XXE Injection via XMLToolMessagelangroid
Apr 202518 advisories
- HighData exposure via ZeroMQ on multi-node vLLM deploymentvllm
- MediumTransformers Regular Expression Denial of Service (ReDoS) vulnerabilitytransformers
- CriticalvLLM Vulnerable to Remote Code Execution via Mooncake Integrationvllm
- MediumvLLM: Quadratic Time Complexity in Input Token Processing​ leads to denial of servicevllm
- Mediumn8n Vulnerable to Stored XSS through Attachments View Endpointn8n
- CriticalCVE-2025-24357 Malicious model remote code execution fix bypass with PyTorch < 2.6.0vllm
- CriticalPyTorch: `torch.load` with `weights_only=True` leads to remote code executiontorch
- MediumPyTorch Improper Resource Shutdown or Release vulnerabilitytorch
- MediumvLLM vulnerable to Denial of Service by abusing xgrammar cachevllm
- CriticalBentoML's runner server Vulnerable to Remote Code Execution (RCE) via Insecure Deserializationbentoml
- HighFlowise Vulnerable to SQL Injection via `tableName` Parameterflowise-components
- HighPicklescan Vulnerable to Exfiltration via DNS via linecache and ssl.get_server_certificatepicklescan
- HighFlowiseDB vulnerable to SQL Injection by authenticated usersflowise
- MediumPicklescan failed to detect to some unsafe global function in Numpy librarypicklescan
- MediumPicklescan missing detection when calling built-in python library function timeit.timeit()picklescan
- CriticalBentoML Allows Remote Code Execution (RCE) via Insecure Deserializationbentoml
- MediumLMDeploy Improper Input Validation Vulnerabilitylmdeploy
- MediumInternLM LMDeploy code injection vulnerabilitylmdeploy
Mar 202593 advisories
- MediumPyTorch is Vulnerable to Memory Consumption through pad_packed_sequence Functiontorch
- LowPyTorch is vulnerable to memory corruption through its torch.lstm_cell functiontorch
- LowPyTorch is vulnerable to memory corruption through its torch.jit.script functiontorch
- MediumPyTorch is vulnerable to memory corruption through its unpack_sequence functiontorch
- LowPyTorch susceptible to local Denial of Servicetorch
- HighOllama Divide by Zero Vulnerabilitygithub.com/ollama/ollama
- HighOpen WebUI Vulnerable to a Session Fixation Attackopen-webui
- HighH2O Vulnerable to Arbitrary File Overwrite via File Exporth2o
- MediumMLflow Uncontrolled Resource Consumption vulnerabilitymlflow
- CriticalPyTorch Lightning path traversal vulnerabilitylightning, pytorch-lightning
- LowMLflow has Weak Password Requirementsmlflow
- HighMLflow has a Local File Read/Path Traversal in dbfsmlflow
- HighLiteLLM Vulnerable to Remote Code Execution (RCE)litellm
- MediumBentoML Open Redirect vulnerabilitybentoml
- HighH2O Vulnerable to Denial of Service (DoS) via `HEAD` Requesth2o
- HighOpen WebUI Unauthenticated Multipart Boundary Denial of Service (DoS) Vulnerabilityopen-webui, open-webui
- Mediumlangchain-core allows unauthorized users to read arbitrary files from the host file systemlangchain-core, langchain-core, langchain-core
- HighGradio DOS in multipart boundry while uploading the filegradio
- HighOpen WebUI denial of service through endpoint for converting markdownopen-webui
- CriticalvLLM Deserialization of Untrusted Data vulnerabilityvllm
- HighZenML unauthenticated DoS via Multipart Boundryzenml
- MediumAgentScope stored cross-site scripting (XSS) vulnerabilityagentscope
- MediumTransformers Regular Expression Denial of Service (ReDoS) vulnerabilitytransformers
- HighAgentScope directory traversal vulnerability in /read-examplesagentscope
- HighH2O Vulnerable to Denial of Service (DoS) via Large GZIP Parsingh2o
- HighOpen WebUI Unauthenticated Multipart Boundary Denial of Service (DoS) Vulnerability in api/chat/fileopen-webui
- HighAgentScope Cross-Origin Resource Sharing (CORS) vulnerabilityagentscope
- HighH2O Vulnerable to Denial of Service (DoS) via `/3/ParseSetup` Endpointh2o
- MediumGradio Vulnerable to Open Redirectgradio
- HighGradio Vulnerable to Denial of Service (DoS) via Crafted Zip Bombgradio
- HighOpen WebUI Cross-Site Request Forgery (CSRF) Vulnerabilityopen-webui
- HighLiteLLM Has a Leakage of Langfuse API Keyslitellm
- HighOllama Allows Out-of-Bounds Readgithub.com/ollama/ollama
- MediumMLflow Cross-Site Request Forgery (CSRF) vulnerabilitymlflow
- HighPyTorch Lightning denial of service vulnerabilitypytorch-lightning
- CriticalBentoML deserialization vulnerabilitybentoml
- HighOllama Divide By Zero vulnerabilitygithub.com/ollama/ollama
- HighOpen WebUI lacks authentication for the `api/v1/utils/pdf` endpointopen-webui
- CriticalAgentScope Deserialization Vulnerabilityagentscope
- CriticalAgentScope path traversal vulnerabilityagentscope
- MediumOpen WebUI Has Improper Access Control Leading to Arbitrary Prompt Readopen-webui
- HighOpen WebUI Uncontrolled Resource Consumption vulnerabilityopen-webui, open-webui
- CriticalvLLM allows Remote Code Execution by Pickle Deserialization via AsyncEngineRPCServer() RPC server entrypointsvllm
- HighAgentScope Path Traversal in /api/fileagentscope
- HighOllama Allocation of Resources Without Limits or Throttling vulnerabilitygithub.com/ollama/ollama
- HighOpen WebUI allows Remote Code Execution via Arbitrary File Upload to /audio/api/v1/transcriptionsopen-webui
- HighLiteLLM Vulnerable to Denial of Service (DoS) via Crafted HTTP Requestlitellm
- HighLiteLLM Has an Improper Authorization Vulnerabilitylitellm
- HighOpen WebUI Uncontrolled Resource Consumption vulnerabilityopen-webui, open-webui
- HighH2O Vulnerable to Arbitrary File Overwriteh2o
- HighLiteLLM Reveals Portion of API Key via a Logging Filelitellm
- HighOpen WebUI stored cross-site scripting (XSS) vulnerabilityopen-webui
- MediumOpen WebUI Allows Viewing of Admin Detailsopen-webui
- HighLiteLLM Vulnerable to Denial of Service (DoS)litellm
- HighOpen Neural Network Exchange (ONNX) Path Traversal Vulnerabilityonnx
- CriticalH2O Deserialization of Untrusted Data Vulnerabilityh2o
- HighBentoML vulnerable to Uncontrolled Resource Consumptionbentoml
- HighBentoML Denial of Service (DoS) via Multipart Boundarybentoml
- MediumOpen WebUI Vulnerable to Cross-Site Scripting (XSS) via Chat File Uploadopen-webui
- HighLlamaIndex Improper Handling of Exceptional Conditions vulnerabilityllama-index-core
- CriticalAgentScope path traversal vulnerability in save-workflowagentscope
- HighLlamaIndex vulnerable to Creation of Temporary File in Directory with Insecure Permissionsllama-index
- HighOpen WebUI Allows Arbitrary File Reading and Deletionopen-webui
- MediumLlamaIndex Uncontrolled Resource Consumption vulnerabilityllama-index
- MediumH2O Vulnerable to Execution of Arbitrary Filesh2o
- HighH2O Vulnerable to Denial of Service (DoS) via `/3/ImportFiles` Endpointh2o
- HighOllama Denial of Service (DoS) via Null Pointer Dereferencegithub.com/ollama/ollama
- MediumOpen WebUI Vulnerable to Cross-Site Request Forgery (CSRF)open-webui
- HighAgentScope arbitrary file download vulnerability in rpc_agent_clientagentscope
- HighGradio Vulnerable to Arbitrary File Deletiongradio
- CriticalvLLM deserialization vulnerability in vllm.distributed.GroupCoordinator.recv_objectvllm
- MediumGradio Path Traversal vulnerabilitygradio
- MediumMLflow Uncontrolled Resource Consumption vulnerabilitymlflow
- HighGradio Vulnerable to Denial of Service (DoS) via Crafted HTTP Requestgradio
- HighOllama Vulnerable to Denial of Service (DoS) via Crafted GZIPgithub.com/ollama/ollama
- HighOpen WebUI has vulnerable dependency on starlette via fastapiopen-webui
- MediumLocalAI Cross-Site Scripting (XSS) vulnerability in its search functionalitygithub.com/mudler/LocalAI
- HighOpen WebUI Uncontrolled Resource Consumption vulnerabilityopen-webui
- HighH2O Vulnerable to Denial of Service (DoS) and File Writeh2o
- HighH2O Vulnerable to Denial of Service (DoS) via `/3/Parse` Endpointh2o
- MediumvLLM denial of service via outlines unbounded cache on diskvllm
- CriticalvLLM Allows Remote Code Execution via Mooncake Integrationvllm
- CriticalFlowise allows arbitrary file write to RCEflowise
- CriticalFlowise Pre-auth Arbitrary File Uploadflowise
- HighArbitrary Code Execution via Crafted Keras Config for Model Loadingkeras
- MediumZip Exploit Crashes Picklescan But Not PyTorchpicklescan
- LowPyTorch Tuple Handler is Vulnerable to Memory Corruption through Manipulation of None Argumenttorch
- MediumZip Flag Bit Exploit Crashes Picklescan But Not PyTorchpicklescan
- LowPyTorch: Manipulation of the argument scale/zero_point leads to improper initialization via Quantized Sigmoid Moduletorch
- Mediumray vulnerable to Insertion of Sensitive Information into Log Fileray
- HighFlowiseAI Flowise arbitrary file upload vulnerabilityflowise
- MediumPicklescan Allows Remote Code Execution via Malicious Pickle File Bypassing Static Analysispicklescan
- MediumPyTorch Model Files Can Bypass Pickle Scanners via Unexpected Pickle Extensionspicklescan
Feb 20252 advisories
Jan 20253 advisories
Dec 20241 advisories
Nov 202412 advisories
- High@lobehub/chat Server Side Request Forgery vulnerability@lobehub/chat
- HighMLflow's excessive directory permissions allow local privilege escalationmlflow
- HighDeserialization of Untrusted Data in Hugging Face Transformerstransformers
- HighDeserialization of Untrusted Data in Hugging Face Transformerstransformers
- HighDeserialization of Untrusted Data in Hugging Face Transformerstransformers
- HighFlowise OverrideConfig security vulnerabilityflowise
- MediumMissing ratelimit on passwrod resets in zenmlzenml
- MediumGradio vulnerable to arbitrary file read with File and UploadButton componentsgradio
- Mediumgradio Server Side Request Forgery vulnerabilitygradio
- MediumLangflow vulnerable to remote code executionlangflow
- HighAgentScope uses `eval`agentscope
- LowLocalAI Cross-site Scripting vulnerabilitygithub.com/mudler/LocalAI
Advisories published before November 2024, mostly for agent frameworks, are listed on each product's page.
Sources: GitHub Advisory Database, reviewed and malware advisories (CC BY 4.0, https://github.com/advisories); advisories published by Cursor (https://github.com/cursor/cursor/security/advisories); advisories published by OpenClaw (https://github.com/openclaw/openclaw/security/advisories); advisories published by vLLM (https://github.com/vllm-project/vllm/security/advisories); advisories published by Open WebUI (https://github.com/open-webui/open-webui/security/advisories); advisories published by n8n (https://github.com/n8n-io/n8n/security/advisories); advisories published by Flowise (https://github.com/FlowiseAI/Flowise/security/advisories); advisories published by Langflow (https://github.com/langflow-ai/langflow/security/advisories); advisories published by the MCP Python SDK project (https://github.com/modelcontextprotocol/python-sdk/security/advisories); advisories published by MLflow (https://github.com/mlflow/mlflow/security/advisories); CISA Known Exploited Vulnerabilities catalog (CC0). Data as of 2026-11-22. ShieldMap describes how each problem works and how to fix it; it does not publish exploit code. Report a correction to [email protected].