proj-fix

Pass

Audited by Gen Agent Trust Hub on Sep 23, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data which is a standard part of its functional scope.
  • Ingestion points: The skill ingests user-provided "Problem descriptions" and "Error logs" in Step 1 (SKILL.md).
  • Capability inventory: The skill facilitates file identification, code modification (Step 4), and Git operations.
  • Boundary markers: The instructions do not define specific delimiters or guardrails to differentiate between user data and instructions within the logs.
  • Sanitization: There are no explicit steps for sanitizing or escaping the content of ingested error logs.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 23, 2026, 10:08 AM
Security Audit — agent-trust-hub — proj-fix