The Agent Skills Directory

[SAFE]: The skill is designed as a collaborative planning tool that enforces a hard constraint against code execution or implementation until a design has been presented and approved by the user. It limits its activities to documentation and git operations within the local project directory.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through its requirement to explore project context. 1. Ingestion points: Project files, documentation, and recent git commits. 2. Boundary markers: Absent. 3. Capability inventory: File system reading, file system writing (restricted to docs/plans/ folder), and git commit commands. 4. Sanitization: No explicit sanitization or filtering of ingested file content. Despite this surface, the risk is classified as safe due to the mandatory human-in-the-loop approval mechanism that prevents the agent from transitioning to implementation skills without verification.

brainstorming