The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process experimental skill files from a sandbox directory for promotion to a canonical bundle. * Ingestion points: Sandbox skill directories, validation reports, and associated documentation files (SKILL.md, references, scripts). * Boundary markers: The skill specifies that experimental skills must have clear boundaries and pass validation, though it lacks explicit prompt-level delimiters for the ingestion process. * Capability inventory: File system operations including directory creation, copying files (cp), creating symlinks (ln -s), updating project configuration (skills-lock.json), and modifying documentation (provenance.md). * Sanitization: Relies on external validation tools (check-all.mjs and specific validation scripts) to verify skill integrity before and after promotion.
[COMMAND_EXECUTION]: Instructs the agent to execute local validation utilities to ensure the integrity of the skill bundle. * Evidence: Steps 1, 3, and 5 require running check-all.mjs and other validation scripts to verify the state of the sandbox and the canonical bundle. These are executed within the local project context.

skill-promoter