audit-decisions

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data sources including lane reports, transcripts, diffs, and commits to reconstruct decision history, which could contain malicious instructions designed to manipulate the auditing agent.
  • Ingestion points: The skill explicitly instructs the agent to read lane reports, session history, transcripts, and diffs to trace back decisions.
  • Boundary markers: The instructions lack explicit requirements for delimiters or 'ignore instructions' warnings when parsing external history or diffs.
  • Capability inventory: The agent has the capability to write to local files (e.g., docs/notes/<branch>-decisions.md) and update Pull Request bodies or comments.
  • Sanitization: There is no mention of sanitizing or validating external content before it is interpolated into the audit report or ledger.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 02:27 AM