plan-secrets-audit

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by scanning repository files and git history, which represents a surface area where instructions embedded in that data could attempt to influence the AI's behavior.\n
  • Ingestion points: Repository working tree files and git commit history.\n
  • Boundary markers: The skill includes 'Guardrails' and 'Self-critique' sections that enforce a 'plan-only' mode and prevent autonomous execution.\n
  • Capability inventory: File content scanning, credential classification, and report generation in markdown format.\n
  • Sanitization: Instructions strictly limit the output of any detected sensitive values to the last four characters only.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:27 PM