pilot
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill implements a defensive and transparent development process. No indicators of credential theft, remote code execution, or persistence were found. The instructions emphasize diff inspection and evidence-based reporting.
- [INDIRECT_PROMPT_INJECTION]: The skill processes implementation plans and repository data, representing a surface for indirect prompt injection. The risk is minimized by the skill's core architecture, which mandates diff reviews and verification gates (e.g., referee review, tests) before any change is finalized. Ingestion points: Plan files (docs/plans/*.md) and repository code. Boundary markers: None specified. Capability inventory: File system modification and local command execution. Sanitization: Manual and automated verification loops.
- [COMMAND_EXECUTION]: The skill invokes local development tools for testing and building. These commands are repository-specific and necessary for the stated purpose of code implementation and verification.
Audit Metadata