commit
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes several git commands including
git status,git diff,git add, andgit commit. These operations are standard for the skill's purpose and are executed within the local repository context. - [PROMPT_INJECTION]: The skill reads repository data (git diffs and status) which serves as a surface for indirect prompt injection if an attacker places malicious instructions in code comments. This is mitigated by a human-in-the-loop requirement for all state-changing actions.
- Ingestion points: Results from
git diffandgit statuscommands defined inSKILL.md. - Boundary markers: No specific delimiters are specified for separating diff data from instructions.
- Capability inventory: The skill can stage files (
git add), apply cached patches (git apply), and create commits (git commit). - Sanitization: The workflow includes an explicit step to flag secrets,
.envfiles, and credentials to prevent accidental staging.
Audit Metadata