llm-eval-type-selector

Pass

Audited by Gen Agent Trust Hub on Apr 23, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read context files from the local workspace (e.g., CLAUDE.md, product-marketing-context.md) and user-provided failure modes. While these represent untrusted data sources, the skill only uses the information to generate textual recommendations for evaluation strategies, presenting minimal security risk.
  • Ingestion points: Local workspace documentation files and user-provided failure descriptions.
  • Boundary markers: Not explicitly defined for external file content.
  • Capability inventory: The skill only produces text-based recommendations and does not invoke external tools or execute code.
  • Sanitization: No specific sanitization or escaping of input data is mentioned.
  • [DATA_EXPOSURE]: The skill requests access to documentation files within the user's project to tailor its advice. This is routine behavior for a development-focused agent and does not involve accessing sensitive system files or credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 23, 2026, 01:21 PM
Security Audit — agent-trust-hub — llm-eval-type-selector