The Agent Skills Directory

[SAFE]: The skill serves a defensive purpose by providing a framework for auditing CI/CD configurations against SLSA and OWASP standards. No malicious code or command execution patterns were found.- [PROMPT_INJECTION]: Although automated scanners detected keywords like 'ignore previous instructions', these appear within a defensive safety notice. This notice instructs the agent to treat such phrases as data when encountered in analyzed files, rather than following them as instructions.- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted external data, which is an inherent risk factor. Evidence Chain: 1. Ingestion points: CI/CD configuration files (e.g., .github/workflows/*.yml) located via Glob and Read tools. 2. Boundary markers: The skill instructs the agent to adhere to a structured report template and treat analyzed content as data. 3. Capability inventory: Uses Read, Grep, and Glob tools for analysis. 4. Sanitization: Explicit instructions are provided to ignore embedded instructions and redact sensitive data.- [DATA_EXFILTRATION]: The skill does not have network access (no curl/wget) and explicitly directs the agent to redact secrets discovered during the audit process to prevent accidental exposure.

pipeline-security