The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the official GitHub CLI (gh label, gh api, gh pr) to interact with repository metadata. These commands are scoped to the specific task of managing labels and verifying GitHub Actions integrity.
[EXTERNAL_DOWNLOADS]: The skill utilizes several GitHub Actions to automate label synchronization and PR labeling. All actions are from trusted or well-known sources (GitHub's own actions organization and crazy-max), and the provided templates use SHA-pinning to ensure supply chain integrity.
[REMOTE_CODE_EXECUTION]: While the skill sets up GitHub Workflows that execute code in a CI environment, the risk is mitigated by using verified, SHA-pinned actions. The actions/labeler workflow uses the pull_request_target trigger, but it is safely configured to only run the labeler logic with minimal permissions (contents: read, pull-requests: write), avoiding the execution of arbitrary code from pull requests.
[INDIRECT_PROMPT_INJECTION]: The skill defines an ingestion surface for repository configuration files (.github/labels.yml). To mitigate the risk of malicious instructions within these files, the skill includes explicit instructions for the agent to diff changes and require human-in-the-loop authorization for any destructive actions, particularly label deletions.
[SAFE]: The skill adheres to the principle of least privilege in its requested workflow permissions and provides clear verification steps for the user to validate the state of the repository after execution.

labels-as-code