clawpathy-autoresearch

To automate SKILL.md file tuning for LLM agents, this Skill uses an LLM to iteratively rewrite the file and judge its quality against a specific task rubric. It eliminates manual iteration by applying eval-driven improvements to enhance performance.

Eval-driven skill development uses an LLM-as-judge mechanism to evaluate and iteratively rewrite SKILL.md files. This process ensures the LLM agent performs better against a specific task rubric by automating quality assessments and continuous improvements.

Yes, you need Python, Claude, and Opus to execute the SKILL.md files and judge their quality. These dependencies are required to run the automated skill tuning process and evaluate the LLM agent iterations.

Yes, you can use an LLM-as-judge to iteratively rewrite agent SKILL.md files. The tool evaluates each version against a rubric to automatically improve the file for better LLM performance on text analysis or other tasks.

The limitations of using an automated LLM judge for skill development include the dependency on Python and Opus to execute files. The tuning quality is bounded by the defined rubric, requiring a clear task definition to achieve effective eval-driven improvements.