The Agent Skills Directory

[SAFE]: The skill consists of instructional content and reference documentation for auditing LLM evaluation methodologies. No malicious code, data exfiltration patterns, or obfuscation techniques were identified.
[EXTERNAL_DOWNLOADS]: The skill references official documentation for well-known LLM evaluation frameworks (Promptfoo, DeepEval, Braintrust, LangSmith) and academic citations (arXiv, ACL Anthology). These are all reputable and trusted sources.
[PROMPT_INJECTION]: The skill instructs an agent to process external configuration and test data for auditing purposes. While this creates a data ingestion surface, the skill provides structural audit criteria that help the agent maintain a focused analysis role rather than executing arbitrary content from analyzed files.

llm-eval-anti-patterns