harden-github-actions
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the zizmor static analysis tool using
uvx zizmor,pipx run zizmor, or by pulling a Docker image fromghcr.io/zizmorcore/zizmor. These are standard methods for running specialized security tools in a development environment.\n- [REMOTE_CODE_EXECUTION]: The tool is executed via package runners likeuvxorpipx, or as a Docker container. These execution methods are used to run a specific security utility without global installation.\n- [COMMAND_EXECUTION]: The skill uses Bash to execute auditing commands and the GitHub CLI (gh api) to resolve Git references. These operations are restricted to auditing local workflow files and querying the official GitHub API.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process.github/workflows/*.ymlfiles, which could potentially contain malicious instructions if sourced from untrusted templates.\n - Ingestion points: Workflow files in the
.github/workflows/directory.\n - Boundary markers: The skill uses a dedicated static analysis tool (zizmor) rather than direct LLM parsing for the security audit.\n
- Capability inventory: The skill has access to Bash for running the analyzer, as well as file read/write capabilities to fix findings.\n
- Sanitization: The skill relies on the zizmor tool's internal parsing logic.
Audit Metadata