commit-expert
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository diffs and working tree states, creating a potential surface for indirect prompt injection. This is mitigated by the skill's strict requirement for human oversight.\n
- Ingestion points: Repository diffs, status outputs, and local policy configuration files.\n
- Boundary markers: Extensive instructions throughout the skill mandate explicit user authorization before any mutation and require manual verification of staged changes.\n
- Capability inventory: Filesystem write access via Git staging and committing, and network access via Git push operations.\n
- Sanitization: Includes a dedicated 'Secret and sensitive content check' logic in references/clean-tree-checklist.md to prevent credential leakage and employs temporary files for commit messages to prevent shell injection.\n- [DYNAMIC_EXECUTION]: The skill generates temporary files for handling commit messages, which is a defensive engineering practice to avoid shell-based injection vulnerabilities.\n
- Evidence: Found in references/commit-execution.md, utilizing
mktempandgit commit -Fto ensure multi-line messages are handled securely by the underlying operating system.
Audit Metadata