The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Node.js scripts for operational tasks, including .agents/scripts/precheck.mjs for environment validation, .agents/scripts/cooking.mjs for status reporting, and .agents/scripts/spec-files.mjs for parsing and validating the generated specification files. These are documented internal utility scripts.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where requirements provided by the user or extracted from goal.md are interpolated into a specification template. While this constitutes an injection surface, the impact is limited to the generated markdown documentation. Ingestion points: User-provided requirement descriptions and goal.md files. Boundary markers: The skill lacks explicit 'ignore embedded instructions' markers, though it follows a strict template. Capability inventory: The skill has permissions to read internal project documentation (e.g., ARCHITECTURE.md, CODE-MAP.md, PROJECT.md), execute local Node.js validation scripts, and write to the .agents/cooking/ directory. Sanitization: No specific sanitization or filtering logic is defined for the requirement strings before they are written to the spec.md file.

to-spec