The Agent Skills Directory

[SAFE]: The skill defines a structured workflow for requirements gathering and design documentation.
[SAFE]: It implements a human-in-the-loop control mechanism that requires user approval of design sections before any implementation actions are taken.
[PROMPT_INJECTION]: The skill ingests untrusted data when reading project files and commit history to establish context. While this presents an indirect prompt injection surface, the design's focus on user-approved documentation and the absence of high-privilege tool calls mitigates the risk. 1. Ingestion points: Instructions to explore project context by checking files, docs, and recent commits in SKILL.md. 2. Boundary markers: None explicitly defined for external content ingestion. 3. Capability inventory: File writing to docs/plans/, git commit execution, and invocation of the writing-plans skill. 4. Sanitization: No automated sanitization of project file content is specified.

brainstorming