The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) to perform repository management tasks such as creating labels, managing project boards, and linking repositories to projects.
[EXTERNAL_DOWNLOADS]: The skill references resources from the author's own repository (fellowship-dev/dogfooded-skills) and configures workflows that utilize established GitHub Actions from trusted organizations like actions/checkout and anthropics/claude-code-action.
[INDIRECT_PROMPT_INJECTION]: The skill installs several GitHub Actions templates (e.g., claude-code-review.yml, claude-issue-triage.yml) that process potentially untrusted external content.
Ingestion points: Workflow files located in .github/workflows/ that ingest pull request diffs, descriptions, and issue comments.
Boundary markers: The templates do not include explicit delimiter markers or specific instructions to the underlying model to ignore instructions embedded within the data being reviewed.
Capability inventory: The installed workflows possess the capability to write to the repository's PRs and issues via the GitHub CLI, including editing labels and posting comments.
Sanitization: There is no evidence of specific sanitization or filtering of the external data before it is processed by the AI-powered action.

setup-github