commit
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository (diffs, commit history, branch names) which could contain embedded instructions. • Ingestion points: Git repository state accessed via git status, git diff, and git log (SKILL.md). • Boundary markers: No specific delimiters are used when interpolating file contents or history into the prompt context. • Capability inventory: The skill can execute git commit, git add, and other shell-based Git commands (SKILL.md). • Sanitization: No explicit sanitization or escaping of the ingested Git data is performed before processing.
- [SAFE]: The skill includes explicit security guardrails instructing the agent to scan for and report credentials or private keys before they are committed to the repository.
Audit Metadata