evals-write-spec
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: No malicious patterns detected. The skill acts as a documentation resource for developers to author and run evaluation tests for AI models.
- [NO_CODE]: The skill does not provide executable scripts or automated tasks; it provides instructional content and code snippets intended for manual implementation by developers.
- [INDIRECT_PROMPT_INJECTION]: While the evaluation framework described inherently processes untrusted data as 'datasets' (ingestion points in SKILL.md), this behavior is the primary purpose of an evaluation tool. The risk is minimized as the skill provides templates for developer use rather than automated execution with elevated privileges. Capability inventory includes internal Elastic fixtures such as 'fetch', 'esClient', and 'executorClient'.
Audit Metadata