The Agent Skills Directory

[SAFE]: The skill serves as a comprehensive guide for GitHub Actions and does not contain malicious code or instructions.
[SAFE]: It emphasizes security best practices, such as the use of encrypted GitHub Secrets for sensitive information like API keys and tokens, and pinning action versions.
[EXTERNAL_DOWNLOADS]: Mentions numerous official and well-known GitHub Actions (e.g., from actions/, docker/, aws-actions/, github/, vercel/). These are standard, trusted components for workflow automation.
[COMMAND_EXECUTION]: Provides examples of shell commands used within CI/CD pipelines, including package management (npm, pip, uv), build tools (docker), and cloud CLIs (aws). These are legitimate within the context of the skill's intended purpose.
[COMMAND_EXECUTION]: Includes the use of mxschmitt/action-tmate for interactive debugging of failed jobs, which is a standard tool providing shell access in CI environments for troubleshooting purposes.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface where workflows ingest untrusted data from repository code or pull requests.
Ingestion points: Workflow execution involves processing repository code and event metadata (SKILL.md).
Boundary markers: Recommendations include environment protection rules and branch protection with required checks (SKILL.md).
Capability inventory: CI jobs utilize shell execution (run) and various network-enabled deployment actions (SKILL.md).
Sanitization: The skill explicitly documents security scanning using CodeQL and Trivy to detect vulnerabilities in code and containers (SKILL.md).

github-actions