The Agent Skills Directory

[Command Execution]: The skill utilizes the GitHub CLI (gh) and the zizmor scanner to retrieve and analyze workflow files. These operations are restricted to read-only access, aligning with the skill's stated purpose of auditing security without modifying code.
[Indirect Prompt Injection Surface]: Since the skill audits external GitHub Actions workflows, it processes untrusted content. This represents a potential indirect prompt injection surface (Ingestion points: gh repo clone, zizmor github:<repo>, gh api in SKILL.md). The risk is addressed through explicit boundary markers (Golden rule 3) that instruct the agent to treat all external content strictly as data. The capability inventory is limited to analysis and reporting tools (gh CLI, zizmor, python3 in SKILL.md), with no automated writing capabilities, and sanitization relies on the agent's adherence to these safety directives.
[Dynamic Execution]: The skill includes a pre-flight check that executes a local Python module (setup_preflight) to verify environmental configuration. This is a standard procedure for ensuring all necessary components are correctly set up before starting a scan.
[External Dependency Management]: The skill recommends the installation of the zizmor package via established package managers. This dependency is central to the skill's functionality as a security auditor and is referenced from well-known sources.

workflow-security-audit