optimize-agent-instructions
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests content from repository instruction files that could contain untrusted instructions. This is a core function of the auditing process, and the skill includes specific logic to detect and strip foreign or stale content.
- Ingestion points: CLAUDE.md, GEMINI.md, and .github/copilot-instructions.md.
- Boundary markers: Absent.
- Capability inventory: File reading (Read), file writing (Write), and file system listing (Bash).
- Sanitization: The skill contains an explicit audit checklist designed to identify and remove stale artifacts and foreign rules.
- [EXTERNAL_DOWNLOADS]: The skill references an external GitHub repository as the source for behavioral principles and attribution. This reference is informational and does not involve automated execution, script downloading, or package installation.
Audit Metadata