git-commit-preparation
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill executes standard version control commands (
git status,git diff,git commit,git push) and GitHub CLI operations (creating and merging pull requests). It also invokes a local Python quality gate script (python3 tools/quality_gate.py) as part of its verification process. - [DATA_EXFILTRATION]: No malicious data exfiltration was detected. The skill explicitly forbids staging or committing credentials, tokens, private configuration, or local databases, and it restricts network operations to pushing code to the repository's configured origin.
- [PROMPT_INJECTION]: The instructions do not contain any patterns intended to bypass AI safety filters. Instead, the skill enforces strict behavioral constraints, instructing the agent to never bypass verification, never push to the main branch directly, and to stop immediately if security risks like sensitive data are detected in diffs.
- [REMOTE_CODE_EXECUTION]: No remote script downloads or execution patterns were found. Dependency verification and quality checks are routed through local tools and well-known services like SonarQube.
- [SAFE]: The skill implements significant safety mechanisms, including mandatory branch handling rules, file classification requirements, and multi-phase review cycles (Phase 5 and Phase 9) to ensure only intended, non-sensitive changes are published.
Audit Metadata