agentseal

What the README says you get

• Other
  Local agent config scanningScans all AI agent configurations on your machine. No API key, no network calls
• Other
  Prompt red-teamingTests a system prompt against 225 adversarial attack probes
• Other
  MCP poisoning detectionaudits its tool descriptions for poisoning
• Other
  Real-time file monitoringWatches agent config files in real time, alerts on threats, quarantines payloads
• Other
  Semantic attack analysisembedding similarity (MiniLM-L6-v2) catches rephrased attacks that bypass patterns
• Other
  MCP registry trust scoreslive trust scores from the MCP Security Registry (6,600+ servers)
• Other
  Custom YAML rulesYAML rules to enforce org-specific policies
• Other
  Python APIPython API
• Other
  TypeScript APIprogrammatic TypeScript API
• Other
  Local LLM supportuse Ollama for free local inference
• Testing
  CI/CD SARIF outputUse `--output sarif` for GitHub Security tab integration

Extracted from the kit's README by a local AI model (deepseek-flash) on Sep 29, 2026. Every item is checked against an exact README quote; 3 unverifiable items were discarded. A README claim is the maker's description, not something we tested.