commit
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell commands including
git status,git diff,git log,git add,git commit, andgit push. These operations are standard and necessary for the skill's stated purpose of managing a Git workflow.- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from the local repository state. - Ingestion points:
git status,git diff, andgit logoutput inSKILL.md. - Boundary markers: External data is wrapped in markdown code blocks.
- Capability inventory: The skill has the capability to modify the repository (
git add,git commit) and sync with remote servers (git push). - Sanitization: The risk is mitigated by explicit instructions requiring conventional commit formats and imperative mood. Furthermore, the skill enforces project safety standards by prohibiting the use of
--no-verifyand preventing the suppression of TypeScript or linting errors, ensuring the agent cannot be easily coerced into committing unsafe code.- [SAFE]: No malicious patterns such as obfuscation, unauthorized data exfiltration, or persistence mechanisms were detected. The skill's behavior is consistent with its documentation and adheres to security best practices for Git management tools.
Audit Metadata