commit
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security risks were identified. The skill instructions focus on standardizing Git operations through conventional commits and organized staging.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection as it processes file diffs and git status output.
- Ingestion points: Reads
git statusoutput and file diffs during the staging process (SKILL.md). - Boundary markers: Instruction #10 acts as a security boundary by requiring the agent to "Ask before staging unrelated user changes, unfamiliar generated churn, or ambiguous dirty tree."
- Capability inventory: The skill uses Git commands (
git status,git add,git commit) via the shell to manage the repository. - Sanitization: No explicit data sanitization is performed, but the workflow is inherently human-in-the-loop, requiring confirmation for ambiguous data.
Audit Metadata