[ PUBLIC · INDEX ]
Public Site Map
Every public ShieldPi page in one place: platform pages, docs, arena, research, blog, account access, and legal resources.
[ Platform ]
HomeShieldPi overview for AI agent security testing, monitoring, and response./Agent WatchtowerRuntime monitoring, endpoint telemetry, forensic timelines, and SOC response./products/agent-watchtowerLeaderboardPublic model security rankings from ShieldPi scan intelligence./leaderboardLive IntelligencePublic intelligence feed, model family trends, and scan methodology signals./intelligenceMethodologyHow ShieldPi runs attacks, judges results, and publishes reproducible evidence./methodology
[ Documentation ]
Docs HomeAPI docs and product guides for scans, monitoring, Guard, webhooks, and SIEM./docsGetting StartedCreate a target, launch a scan, and read the first results./docs/getting-startedAPI ReferenceAuthentication, request shapes, scan endpoints, reports, and public APIs./docs/api-referenceLive Agent MonitorCapture LLM calls, tool use, file access, network activity, and endpoint signals./docs/live-agent-monitorAgent Monitor Deep DiveDetector architecture, alert handling, incident triage, and runtime telemetry./docs/agent-monitorScan ModesBrowser, API, agent, and model scan modes with implementation guidance./docs/scan-modesWebhooksWebhook events, retry behavior, HMAC signing, and integration examples./docs/webhooksSIEM FormatsOCSF, CEF, Syslog, Splunk, Datadog, and Microsoft Sentinel payloads./docs/siem-formatsWatchtower GuardInline block, redact, and inject decisions for LLM agent turns./docs/watchtower-guardWatchtower CommandsAsync response commands for killing sessions, quarantining memory, and disabling tools./docs/watchtower-commands
[ Arena ]
AI Red-Team ArenaCommunity attack lab for validating prompts and feeding Watchtower intelligence./labTodayThe current public challenge target./lab/todayGauntletMulti-stage public arena challenge./lab/gauntletArena LeaderboardCommunity ELO rankings and validated submissions./lab/leaderboardELO SystemHow arena scores and player ratings are calculated./lab/eloVerify SubmissionPublic verification flow for submitted arena attacks./lab/verifyRegisterJoin the AI Red-Team Arena alpha./lab/registerDSA ComplianceArena transparency and moderation information./lab/dsa-compliance
[ Company and Content ]
[ Legal ]
[ Blog Posts ]
Introducing ShieldPi Agent Watchtower — The SOC for AI AgentsWe built the first real-time security monitor for LLM agents. It watches every prompt, tool call, and memory write your AI agents produce — live, across multiple steps and sessions — for attacks that inline guardrails physically can't see. Available now./blog/introducing-agent-watchtowerWhy LLM Security Testing Matters in 2026AI models are deployed everywhere — but most organizations have never tested them for security vulnerabilities. Here's why automated LLM security testing is no longer optional./blog/why-llm-security-mattersTop 10 LLM Vulnerabilities Developers Must KnowA practical walkthrough of the OWASP LLM Top 10 vulnerabilities — with real examples, risk ratings, and mitigation strategies for each./blog/top-10-llm-vulnerabilitiesHow Jailbreak Attacks Work Against AI ModelsAn educational deep-dive into the most common LLM jailbreak techniques: DAN, crescendo, role-play, authority impersonation, and more. Understand the attacks to defend against them./blog/jailbreak-techniques-explainedGPT-4o vs Claude: AI Security ComparisonWe ran 58,000+ attack techniques against both GPT-4o and Claude Sonnet 4.6. Here's how they compare on jailbreaks, prompt injection, data exfiltration, and overall security./blog/gpt4o-vs-claude-securityHow to Secure Your LLM ApplicationPractical steps to secure your LLM deployment: input validation, output filtering, guardrails, monitoring, and automated testing. A hands-on guide for engineering teams./blog/how-to-secure-your-llm
[ Research Papers ]
[ Account Access ]