use-conventional-commits

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill performs git operations (status, diff, add, commit) using standard CLI tools. It employs a shell heredoc with a quoted delimiter ('EOF') to safely handle commit messages, which prevents the shell from interpreting the content of the message as commands.\n- [PROMPT_INJECTION]: The skill handles untrusted file content and diffs to generate summaries, presenting a surface for indirect prompt injection.\n
  • Ingestion points: Diff data from git diff and file contents read directly in Step 3.\n
  • Boundary markers: The skill lacks automated boundary markers but enforces a human-in-the-loop review in Step 5 before committing.\n
  • Capability inventory: git add -A and git commit in Step 6.\n
  • Sanitization: No explicit sanitization of diff content is performed.\n- [DATA_EXFILTRATION]: The skill's scope is limited to local git operations. It explicitly forbids network operations like 'git push', preventing data from leaving the local environment.\n- [SAFE]: The instructions include 'Hard Rules' that prevent common security pitfalls, such as committing credentials or performing destructive history rewrites.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 12:03 AM