skills/wangjq4214/skills/grimoire-map/Gen Agent Trust Hub

grimoire-map

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data as it is designed to analyze third-party repository content.\n
  • Ingestion points: Reads all first-party repository files, manifests, tests, and configurations (SKILL.md).\n
  • Boundary markers: Employs strict JSON schemas (wire-format.md) and instructs the agent to differentiate between observed source facts and inferred behavior to maintain a clear line between data and instructions.\n
  • Capability inventory: Reads repository files, performs directory enumeration, and writes map artifacts to a dedicated .grimoire/map/ directory.\n
  • Sanitization: Contains explicit directives to filter out secrets, credentials, and customer data from all generated artifacts.\n- [COMMAND_EXECUTION]: The skill specifies repository enumeration and validation commands but incorporates path safety rules in wire-format.md that reject absolute paths and parent directory traversal (..), preventing common injection vectors.\n- [DATA_EXPOSURE]: Enforces a strict data minimization policy, ensuring that only concise facts and locators are persisted while sensitive information is excluded.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 11:06 AM