The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No prompt injection or behavior override patterns were detected. The skill uses standard instruction-based keyword matching to serve relevant resources.
[CREDENTIALS_UNSAFE] (SAFE): No hardcoded credentials. The skill explicitly recommends using GitHub Secrets, OIDC, and short-lived tokens instead of long-lived credentials.
[REMOTE_CODE_EXECUTION] (SAFE): While the skill suggests the use of third-party GitHub Actions, it mandates the use of full commit SHAs (SHA pinning) to prevent supply chain attacks, which is a high-standard security practice.
[DATA_EXFILTRATION] (SAFE): No malicious network operations or attempts to exfiltrate sensitive files were found. Network operations mentioned (like npm publish or OIDC auth) are standard CI/CD procedures.
[INDIRECT_PROMPT_INJECTION] (SAFE): The skill demonstrates high awareness of indirect injection through untrusted GitHub event data. It provides specific remediation patterns (using environment variables instead of direct expression interpolation) to mitigate these risks.
[EXTERNAL_DOWNLOADS] (SAFE): All suggested downloads/actions are from well-known providers (Google, AWS, Azure, GitHub Actions) and are accompanied by security warnings regarding their use.

github-action