opensquad-skill-creator

To create Opensquad skills, draft the SKILL.md frontmatter, define a baseline evaluation, run the eval loop, and iterate based on feedback. This workflow automates structured skill creation, testing, and progressive benchmarking improvements.

The evaluation framework provides a repeatable blueprint to design eval prompts, run experiments, and measure outcomes. It establishes clear feedback loops to drive progressive improvements across multiple iterations of skill drafts.

Structuring frontmatter for a new skill involves following enforced validation rules within the creation workflow. This ensures the SKILL.md metadata is correctly formatted before proceeding to the operational body and evaluation phases.

Yes, the workflow automates coordinating evaluation loops by designing eval prompts, running experiments, and interpreting results. This automated benchmarking measures outcomes and drives iterative improvements without manual testing overhead.

The best way to refine skill drafts is layering iterative feedback from evaluation results. This benchmarking approach interprets experiment outcomes to guide progressive improvements, ensuring stronger skills across domains through repeated eval loops.