The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands using gh, git, and jq to retrieve pull request status, manage local branches, and apply formatting/linting tools.
[EXTERNAL_DOWNLOADS]: Communicates with GitHub's REST and GraphQL APIs to fetch pull request metadata, comments, review threads, and CI check statuses.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub pull request comments and review bodies to determine loop progression and exit conditions.
Ingestion points: Metadata and body content from PR comments and review threads fetched from the GitHub API.
Boundary markers: The logic employs filters to exclude bots and the PR author from triggering specific state changes, although the comment text itself is not delimited within internal processing.
Capability inventory: The skill can post comments to GitHub, trigger sub-skills (pr-review, pr-address), and execute local project tooling (e.g., poetry run format, pnpm lint).
Sanitization: Uses jq for structured data parsing and implements regex normalization for user-supplied pull request URLs and identifiers.

pr-polish