The Agent Skills Directory

[SAFE]: The skill describes a legitimate administrative and quality assurance workflow for managing software development phases. No malicious intent or suspicious instructions were found.
[NO_CODE]: The skill contains natural language instructions and process definitions rather than executable scripts or code files, which significantly reduces the technical attack surface.
[PROMPT_INJECTION]: The skill defines ingestion points for untrusted data (phase_artifacts and approval_records), which represents an indirect prompt injection surface. However, the logic is restricted to structural and existence checks rather than executing the contents of these inputs, making the risk negligible.
[DATA_EXFILTRATION]: No network operations, credential harvesting, or sensitive file path access (e.g., SSH keys, environment variables) were identified in the instructions.

gate-validation