llm-as-judge
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown defining a qualitative evaluation process. It does not contain executable scripts, shell commands, or network requests.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and evaluate untrusted 'artifacts' (Phase 3). This creates a surface for indirect prompt injection where a malicious artifact could attempt to manipulate the judging LLM's scores. However, the skill provides a structured rubric and reasoning requirements which serve as natural mitigations.
- Ingestion points: The 'artifact' parameter in the evaluation request structure (SKILL.md).
- Boundary markers: None explicitly defined for the artifact content.
- Capability inventory: No tool execution or system access is granted to the judging process; it is limited to natural language assessment.
- Sanitization: None identified; evaluation is performed on the raw artifact.
Audit Metadata