The Agent Skills Directory

[SAFE]: The skill's executable components are self-contained Node.js scripts that use only built-in modules, avoiding supply chain risks and external dependencies. All file operations are restricted to reading the local documentation directory provided by the user.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided markdown files within the documentation directory, which constitutes an ingestion surface for untrusted data. This is mitigated by the following evidence chain: 1. Ingestion points: The loadDocs function in docs.mjs reads all .md files in the target directory (defaulting to docs/). 2. Boundary markers: Structural requirements such as the index.md listing and specific heading levels are used to define documentation boundaries. 3. Capability inventory: The included Node.js scripts are restricted to local file system read operations (readdir, readFile, stat) only. No network, write, or command execution capabilities are present. 4. Sanitization: The scannableLines function in docs.mjs filters out markdown code blocks and inline code spans to prevent the accidental parsing or execution of example content.

living-project-documentation