The Agent Skills Directory

Command Execution: The skill utilizes local shell commands and scripts to perform documentation tasks such as validation, formatting, and building. These include pnpm run for quality checks and custom Node.js scripts for repository maintenance, which are standard for modern web development workflows.
Evidence: SKILL.md (pnpm run check/build/format), references/information-architecture.md (node tools/directory-entry-ids), and references/information-architecture.md (pnpm exec tsm bin/validate-redirects.ts).
GitHub CLI Interaction: The workflow includes instructions for using the GitHub CLI (gh) to automate pull request reviews and post suggestions. While these commands interact with external APIs, they are performed within the context of the user's authenticated environment and specific pull request tasks.
Evidence: references/reviewing-docs.md provides detailed command templates for gh api and gh pr to manage documentation reviews.
Indirect Prompt Injection Surface: The skill involves reading and analyzing pull request diffs, which are technically untrusted data from a security perspective. This creates a potential surface for indirect prompt injection where malicious instructions could be embedded in the content being reviewed. The skill addresses this by providing a specific set of rules and a structured review process to maintain focus on documentation quality.
Evidence: references/reviewing-docs.md instructs the agent to analyze the full diff of a pull request using gh pr diff to provide constructive feedback.
Sensitive Data Handling: The skill contains explicit instructions to protect sensitive information, warning against the inclusion of secrets or private data in any part of the repository.
Evidence: SKILL.md features a 'Ground rules' section that specifically prohibits putting private information, secrets, or credentials into pages, code, or pull requests.

contributing