model-merging
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references and provides instructions to download established AI research tools from GitHub repositories, specifically from Arcee AI (mergekit), LM-SYS (FastChat), and OpenAI (human-eval).
- [REMOTE_CODE_EXECUTION]: Instructions involve installing the downloaded tools using 'pip install -e .', which executes the setup scripts provided by the developers of these research tools.
- [COMMAND_EXECUTION]: The skill documents the use of command-line interfaces for model manipulation and benchmarking, including 'mergekit-yaml' and Python scripts for generating model responses and calculating scores.
Audit Metadata