build
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a procedural guide for incremental feature development, advocating for test verification and peer review of changes. It contains no commands for privilege escalation, persistence, or data exfiltration.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a data ingestion surface by requiring the agent to read external documentation and upstream repository code to inform implementation.
- Ingestion points: Project files, external API documentation, and pinned upstream source code.
- Boundary markers: Instructions advise staying within the logical task boundary but do not specify explicit prompt delimiters for external content.
- Capability inventory: The workflow involves file reading, file writing (code implementation), git operations (branching and committing), and command execution (running tests).
- Sanitization: The risk is mitigated by the instruction to prioritize authoritative, version-pinned sources over unofficial memory or blog posts.
Audit Metadata