The Agent Skills Directory

[SAFE]: The skill serves as an educational resource for GitHub Actions development. It explicitly identifies security anti-patterns (e.g., hardcoded secrets, outdated versions, and script injection) and provides remediation examples. The hardcoded secrets present in the file are clearly marked as examples of bad practices and use dummy values.
[EXTERNAL_DOWNLOADS]: The skill references documentation and runner image specifications from official GitHub repositories (e.g., github.com/actions/runner-images). These are well-known, trusted sources and do not involve untrusted remote code execution.
[COMMAND_EXECUTION]: The skill utilizes the gh (GitHub CLI) tool via the allowed-tools configuration to fetch release metadata. The suggested commands are limited to read-only metadata retrieval (gh release view) from public repositories, which is appropriate for the skill's purpose of version management.
[INDIRECT_PROMPT_INJECTION]: As the skill involves generating or modifying CI/CD workflows based on user requirements, it possesses an inherent attack surface for indirect prompt injection. However, the skill mitigates this by providing specific templates and instructions on how to use environment variables to prevent shell injection within workflows, adhering to security best practices.

github-actions