OpenClaw Vulnerable to Local File Exfiltration via MCP Tool Result MEDIA: Directive Injection
GHSA-jjgj-cpp9-cvpv · published 2026-03-04
- Severity
- Medium
- CVSS score
- 6.9
- Chance of exploitation (EPSS)
- No CVE
- Fixed in
- 2026.2.21
File exfiltration: Any file within os.tmpdir() (or the OpenClaw state directory) can be read and sent to external messaging channels Secret theft: Temporary files often contain API keys, database credentials, SSH keys, session tokens, and application secrets Cross-application data theft: Other applications' temp files (browser caches, build artifacts, CI/CD secrets) are accessible Silent exfiltration: The file content is sent as a media attachment to messaging channels the attacker can monitor, with no user-visible indication Automated exploitation: If auto-reply is enabled, the malicious tool can be triggered without user interaction
Upgrade openclaw to 2026.2.21 or later. If you start it from an MCP config with npx or uvx, pin that version so an old cached copy is not used.
1. Validate MEDIA: source trust: Only accept MEDIA: directives from OpenClaw's own internal tools (TTS, image generation). Reject or flag MEDIA: directives from external MCP tool results. 2. Remove os.tmpdir() from default localRoots: The temp directory is too broad. Replace with a narrow OpenClaw-specific subdirectory (e.g., path.join(os.tmpdir(), "openclaw-media")). 3. Add source tagging to tool results: Tag each tool result with its source (internal vs. MCP external) and enforce different media access policies for each. 4.
| Package | Registry | Affected | Fixed in |
|---|---|---|---|
| openclaw | npm | < 2026.2.21 | 2026.2.21 |
- CriticalMCP Path Traversal Allows Full-Filesystem Read and JWT Secret Exfiltration Leading to Account Takeoverlangflow, langflow-base
- Critical@zereight/mcp-gitlab: Unauthenticated arbitrary file read via `upload_markdown` enables PAT exfiltration and full account takeover@zereight/mcp-gitlab
- CriticalFlowise RCE via SQL Database Chain Nodeflowise, flowise-components
- CriticalLangroid: SQLChatAgent dangerous-function blocklist can be bypassed with quoted or schema-qualified pg_read_file callslangroid
- Criticalfast-mcp-telegram: Bearer token path traversal bypasses reserved Telegram session protectionfast-mcp-telegram
- CriticalMCP Toolbox HTTP tool path parameters allow traversal to unintended endpointsgithub.com/googleapis/mcp-toolbox
- CriticalOpenClaw: QQBot admin commands could skip DM-only and allowFrom policyopenclaw
- CriticalOpenClaw: Feishu webhook and card-action validation now fail closedopenclaw
- CriticalOpenClaw: Gateway HTTP endpoints re-resolve bearer auth after SecretRef rotationopenclaw
- CriticalOpenClaw: Sandbox escape via TOCTOU race in remote FS bridge readFileopenclaw
- CriticalOpenClaw: Heartbeat context inheritance bypasses sandbox via senderIsOwner escalationopenclaw
- CriticalOpenClaw has a CWD `.env` environment variable injection which bypasses host-env policy and allows config takeoveropenclaw
The impact, patch and workaround text comes from the advisory, shortened (GitHub Advisory Database, CC BY 4.0). Proof-of-concept and reproduction sections and code are left out; the original advisory has them.