The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the GitHub CLI (gh) using execFile to retrieve pull request metadata, comments, and commit status. On macOS, it utilizes the security utility to securely access API keys stored in the system Keychain.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to the official Blocks API at api.blocks.team to manage sessions and fetch review content. The implementation includes strict URL validation to ensure requests are directed to official endpoints.
[INDIRECT_PROMPT_INJECTION]: The logic ingests and parses external pull request comments and review data from GitHub. It employs specific regex patterns and logic to classify review outcomes, with built-in checks to ensure verdicts are dated against the correct commit SHAs to prevent manipulation.
Ingestion points: gh api calls to fetch PR comments and reviews in scripts/blocks-review.mjs.
Boundary markers: The logic lacks explicit delimiters for processing comment bodies but relies on complex regex-based state classification.
Capability inventory: GitHub CLI execution (gh) and API network requests via fetch to the vendor's domain.
Sanitization: Implements strict state classification logic (e.g., isVerdict, isClean, reportsFindings) and validates commit SHAs to ensure data integrity.

blocks