maintain-agent-docs

Pass

Audited by Gen Agent Trust Hub on May 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes git commands to identify code changes that require documentation updates. This is restricted to repository metadata and is necessary for the skill's functionality.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from transcripts and git logs, which represents a surface for indirect prompt injection. This is mitigated by a robust filtering policy that rejects non-durable instructions and sensitive data.
  • Ingestion points: Transcripts from various local sources (Cursor, Codex, Claude) and git repository history as specified in resources/source-discovery.md.
  • Boundary markers: Explicit 'Signal Filter' and 'Confidence Rules' defined in resources/merge-policy.md to identify valid documentation candidates.
  • Capability inventory: File system writes for documentation updates and git command execution for change analysis.
  • Sanitization: Strict prohibition of storing secrets, private data, or raw transcript fragments in documentation or index files.
Audit Metadata
Risk Level
SAFE
Analyzed
May 18, 2026, 11:49 AM
Security Audit — agent-trust-hub — maintain-agent-docs