opensquad-skill-creator

To create and test a new prompt-based skill, use Opensquad Skill Creator to draft the capability, generate a SKILL.md file, and run evaluation loops with test prompts to measure effectiveness before deployment.

The best way to benchmark skill performance is to run evaluation loops that quantify effectiveness using generated test prompts and performance metrics, identifying specific opportunities for iterative improvement.

Yes, you can modify MCP and hybrid skills. Opensquad Skill Creator supports all skill types end-to-end, allowing you to update custom scripts, integrations, and prompt-based capabilities while maintaining performance benchmarks.

Yes, you need pyyaml installed as a dependency. The skill creation and benchmarking process relies on pyyaml to parse configurations and generate the structured YAML front matter required in the SKILL.md output.

An evaluation loop works by executing generated test prompts against the skill, collecting performance metrics, and producing evaluation artifacts that document workflow results to guide the next iteration of modifications.

When skill benchmarking shows no improvements, review the generated evaluation artifacts and test prompts to identify bottlenecks, then modify the SKILL.md workflows and rerun the performance metrics evaluation loop.