The Agent Skills Directory

[SAFE]: The skill is designed as a security-hardening resource for CI/CD pipelines. It explicitly addresses critical vulnerabilities like shell script injection and the risks associated with pull_request_target triggers. All provided examples and patterns use immutable commit SHAs for third-party actions, which is a key security recommendation for production workflows.
[COMMAND_EXECUTION]: The skill performs repository discovery using standard commands (find, ls) to locate existing workflows and identify the technology stack. These operations are restricted to context-gathering within the local project environment.
[EXTERNAL_DOWNLOADS]: The skill references official GitHub Actions and security scanning tools from reputable and trusted organizations, including AWS, Google, Azure, Docker, and StepSecurity. These external resources are used for standard CI/CD functionalities such as cloud authentication, container image management, and egress auditing.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with existing repository workflows to provide context-aware suggestions. While this ingestion surface exists, the skill actively educates the agent on implementing secure patterns to prevent exploitation via untrusted data.
Ingestion points: Repository files in .github/workflows and .github/actions.
Boundary markers: None explicitly defined for the discovery phase.
Capability inventory: Local file discovery via find and ls; agent-level file write and modification capabilities.
Sanitization: The skill provides robust sanitization strategies (e.g., environment variable isolation) for the workflows it helps create.

github-actions-advanced