skill-creator

Automated prompt engineering evaluations and benchmarking require running an evaluation-and-iteration workflow that measures trigger accuracy and response performance. This system supports drafting, evaluating, and iteratively improving skills using bundled scripts and references to enable reproducible optimization.

The best way to set up a craft-and-test loop for skill creation is to draft a goal, create a SKILL.md file with a clear description, run the evaluation loop, and iterate until prompts reliably trigger and outputs meet quality goals.

Yes, you need Python 3 to run eval scripts and validation tooling for skill optimization. The system requires Python 3 to execute the automated evaluation-and-iteration workflow and ensure compatibility and safety.

Yes, you can bundle and organize scripts, references, and assets together for on-demand execution. The system supports using these optional bundled resources to facilitate reproducible iterations during the skill development and evaluation process.

Benchmarking variants helps with triggering behavior optimization by comparing iterations to identify the best description and configuration. This allows you to measure trigger accuracy and response performance across different skill variants to select the most reliable configuration.