llm-consistency-testing
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured prompts and guidelines for analyzing LLM consistency, focusing on reproducibility and evidence-based findings. Analysis of all files shows adherence to safety best practices, including explicit boundaries between facts and inferences.- [NO_CODE]: The repository includes shell and Python scripts within the evals/ directory. These are identified as development-time evaluation fixtures rather than tools intended for execution by the agent during normal skill usage. The scripts utilize standard libraries for string manipulation and table parsing without involving network or sensitive file access.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user scenarios. 1. Ingestion points: User input prompts as described in SKILL.md and prompts/llm-consistency-testing.md. 2. Boundary markers: The prompts/llm-consistency-testing.md file defines a clear Input Audit section and separation between facts and inferences. 3. Capability inventory: No tools are defined in the YAML frontmatter, limiting the agent to text generation. 4. Sanitization: The skill instructs the agent to separate confirmed facts from inferences and use masked data for sensitive work.
Audit Metadata