The Agent Skills Directory

[PROMPT_INJECTION]: Potential surface for indirect prompt injection.
Ingestion points: Fetches live workflow syntax from docs.github.com and reads GitHub Action logs/workflow statuses via the gh CLI.
Boundary markers: Absent; no specific delimiters are used to wrap data retrieved from external documentation or repository logs.
Capability inventory: Scripts have permissions to create and modify GitHub issues, labels, milestones, and projects via PyGithub and gh CLI.
Sanitization: Absent; the skill relies on standard API wrappers without additional sanitization of retrieved text before processing.
[COMMAND_EXECUTION]: Orchestrates GitHub operations via subprocess calls to the gh CLI.
Evidence: github_project_setup.py uses subprocess.run to call gh api graphql with queries constructed via string concatenation. While intended for legitimate project management, this pattern requires careful validation of input parameters.
[EXTERNAL_DOWNLOADS]: References external configuration and documentation from trusted sources.
Evidence: SKILL.md retrieves workflow syntax from GitHub's official documentation and checks Node.js versions from nodejs.org.

github-workflows