The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines automation workflows that execute various shell commands, including git (for commits, tags, and status), find (for file discovery), brew (for dependency installation like pandoc), and gh (for pull request data). It also utilizes custom local scripts such as pdf-to-markdown and validate-conversion.mjs to process documentation.
[EXTERNAL_DOWNLOADS]: The conversion module (CONVERTING.md) uses curl and wget to fetch content from external URLs to convert web documentation into skills. It also suggests downloading the pandoc utility via Homebrew if it is not present on the system.
[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !command`` syntax in PATTERNS.md and SKILL-GUIDE.md to inject real-time project metadata (e.g., git branch, git log, find results) into the agent's context during skill load to provide current environment awareness.
[INDIRECT_PROMPT_INJECTION]: The skill implements a data-ingestion surface that processes untrusted external content (PDFs, EPUBs, and web pages) to generate instruction sets. This risk is mitigated by a multi-phase workflow that requires explicit human review and confirmation (Phase B: User Confirmation) before the final skill is generated and deployed.

authoring-plugins