executing-plans
Pass
Audited by Gen Agent Trust Hub on Apr 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's design prioritizes user oversight and safety. It includes explicit instructions to 'Review critically' any plan before starting and to 'STOP executing immediately' if blockers or gaps are found. It also mandates explicit user consent before modifying primary branches (main/master).
- [SAFE]: The skill processes external implementation plans (e.g., from 'docs/plans/'), which is a potential surface for Indirect Prompt Injection. However, this is managed safely through instructions requiring the agent to identify concerns and raise them with a human partner before execution.
- Ingestion points: Implementation plans located in 'docs/plans/' or provided in the current context.
- Boundary markers: The skill requires a critical review phase and explicitly instructs the agent to 'Ask for clarification rather than guessing'.
- Capability inventory: The agent can read files, execute development tasks, run tests, and perform git operations (PR/merge).
- Sanitization: Safety is maintained through human-in-the-loop checkpoints and reasoning-based verification steps.
Audit Metadata