harness-commit
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill utilizes local Git commands and a validation script for standard repository maintenance. It maintains a strong security posture through an explicit authorization boundary for all staging and committing actions. No external network requests or sensitive data access patterns were identified.
- [INDIRECT_PROMPT_INJECTION]: The skill processes repository state which functions as an untrusted input surface.
- Ingestion points: Git status and diff outputs from the local repository (SKILL.md).
- Boundary markers: The agent instructions lack explicit delimiters to isolate untrusted data during analysis.
- Capability inventory: File system read access via Git tools and execution of local validation scripts.
- Sanitization: The validator script enforces commit message syntax but does not filter or sanitize the underlying repository content.
Audit Metadata