prompt-evaluator
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTIONNO_CODE
Full Analysis
- PROMPT_INJECTION (LOW): Susceptible to Indirect Prompt Injection (Category 8) through ingestion of untrusted data. • Ingestion points: Modes 1, 2, and 3 process external prompt text, files, and user feedback as input for evaluation. • Boundary markers: Absent; the skill lacks explicit delimiters or instructions to treat user-provided data as non-executable text. • Capability inventory: Limited to agent reasoning and structured report generation; no evidence of network operations, file-write, or command execution capabilities in the provided file. • Sanitization: Absent; the skill does not specify any validation or filtering for the content it evaluates, allowing adversarial instructions to potentially influence the analysis outcome or attempt system prompt extraction.
- NO_CODE (LOW): The skill consists entirely of markdown instructions with no executable scripts or binary files, reducing the risk of direct code-based attacks but increasing dependency on the agent's safe interpretation of natural language instructions.
Audit Metadata