Verification Before Completion
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides procedural instructions to improve task reliability and verification accuracy. It does not include any malicious scripts, external connections, or obfuscated content.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs agents to analyze external inputs like VCS diffs and sub-agent reports, creating a surface for potential indirect instructions.
- Ingestion points: The skill explicitly references checking VCS diffs and agent success reports (SKILL.md).
- Boundary markers: No specific delimiters or 'ignore instructions' warnings are provided for these inputs.
- Capability inventory: The agent is expected to execute system commands such as tests, builds, and linters.
- Sanitization: No specific sanitization or filtering logic is mentioned for the ingested data.
Audit Metadata