The Agent Skills Directory

[SAFE]: The skill follows established CI/CD best practices and emphasizes security measures such as least-privilege permissions and secret protection in logs. No malicious commands, obfuscation, or exfiltration patterns were detected.
[EXTERNAL_DOWNLOADS]: The skill templates reference official GitHub Actions from a trusted source (actions/checkout, actions/setup-go, actions/setup-node). These are standard industry components for building workflows.
[INDIRECT_PROMPT_INJECTION]: The skill involves reading repository configuration and package files which are untrusted data. While this presents an attack surface where a malicious repository could attempt to influence the agent's auditing logic, the skill includes decision rules to verify commands and avoid inventing scripts, which mitigates these risks.

cicd-consistency