correct
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of instructional text for repository maintenance and code quality improvements. It does not include any executable scripts, network exfiltration logic, or obfuscated content. References to upstream sources on GitHub are documented for transparency and originate from well-known community repositories.
- [INDIRECT_PROMPT_INJECTION]: The skill instructions involve reading and processing untrusted data from repository history, including commit messages, reverts, and review comments. This ingestion surface is used to identify mistake patterns for architectural prevention. While a surface exists, the instructions guide the agent toward structural repo fixes rather than executing content from the inputs.
- Ingestion points: SKILL.md (reads recent commits, reverts, review comments, and instruction files).
- Boundary markers: Not explicitly defined in the instructions.
- Capability inventory: File system modification within the repository (architecture changes, type definitions, tests).
- Sanitization: Not explicitly defined.
Audit Metadata