The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Bash to perform directory scanning and creation within the documentation root. The tools are restricted to paths matching docs/:*.
[PROMPT_INJECTION]: The skill is designed to read code changes and existing documentation to automate updates, creating an attack surface for indirect prompt injection. Findings include: 1. Ingestion points: The skill reads source code and documentation files via Read, Glob, and Grep tools (SKILL.md Step 1). 2. Boundary markers: No explicit prompt delimiters are mentioned for input data. 3. Capability inventory: Write and Edit tools are used to modify local documentation files (SKILL.md File Access Rules). 4. Sanitization: The skill implements a Claim-Grounding Gate which requires every capability claim in the documentation to be mapped to a specific symbol in the code.
[SAFE]: The skill enforces strict file access rules that limit write operations to documentation directories (docs/, documentation/, doc/, or site/). It also interacts with related author-owned tools like wk-arch-review and wk-learn for workflow consistency.

wk-docs