The Agent Skills Directory

[SAFE]: The skill serves as a technical educational resource and implementation guide for LLM evaluation. No malicious code, prompt injections, or unauthorized data access patterns were found.\n- [EXTERNAL_DOWNLOADS]: The skill references established research papers on arxiv.org and reputable industry blogs. These references are strictly for documentation and research purposes and do not involve executable downloads.\n- [DATA_EXFILTRATION]: No network exfiltration or unauthorized file access was detected. The provided Python scripts perform local calculations for evaluation metrics such as weighted scores and bias indicators.\n- [PROMPT_INJECTION]: The prompt templates provided are specifically structured for evaluation tasks (direct scoring and pairwise comparison) and do not contain instructions to bypass safety guardrails or override system prompts.\n- [REMOTE_CODE_EXECUTION]: No remote code execution or dynamic command execution patterns were found. The Python examples demonstrate logic for data processing and statistical analysis using standard libraries.

advanced-evaluation