The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local bash and Python scripts (e.g., check-all.sh, phase-gate-guard.py, and writing-claim-id-guard.py) to enforce workflow consistency. These are invoked via platform-managed environment variables and do not involve remote code or unsanitized user inputs.
[PROMPT_INJECTION]: The skill processes untrusted content from the project's planning and draft directories. 1. Ingestion points: .planning/PRECIS.md, .planning/OUTLINE.md, and markdown files in the drafts/ directory. 2. Boundary markers: Absent. 3. Capability inventory: The skill utilizes the Bash, Write, Read, and Grep tools, including the execution of the check-all.sh script. 4. Sanitization: Absent. While this constitutes an indirect prompt injection surface, the risk is mitigated by the skill's explicit requirement for human confirmation of all fixes and its instructions for the agent to actively identify AI-specific writing indicators.

writing-validate