commit-workflow
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The skill reads untrusted data from the repository (via git diff and git log) to generate commit messages. This data could contain malicious instructions targeting the AI agent. \n
- Ingestion points: git status, git diff, git log (SKILL.md)\n
- Boundary markers: Absent\n
- Capability inventory: git add, git commit (SKILL.md)\n
- Sanitization: Absent\n- [Command Execution] (SAFE): The skill utilizes standard git commands (status, diff, log, add, commit) to perform its functions. No high-risk command execution or privilege escalation was found.\n- [Data Exposure] (SAFE): The skill includes a specific rule to avoid staging sensitive files such as credentials or .env files, which is a positive security practice.
Audit Metadata