What problem does it solve?
Skill-creator helps you design new AI skills or refine existing ones by turning vague “try it and see” iteration into a structured loop: draft the skill, run trigger-evaluation tests against realistic prompts, review qualitative outputs, then iteratively improve until the skill triggers reliably and performs well.
Core Features & Use Cases
- Skill drafting and improvement loop: Guides you through intent capture, drafting the skill, then iterating based on evaluation results.
- Quantitative trigger evaluation: Runs Claude “trigger/not trigger” tests using a held-out set to reduce overfitting to your prompt set.
- Qualitative review workflow: Launches an eval viewer so you can inspect outputs and leave human feedback.
- Benchmarking and variance analysis: Aggregates pass rates and timing/token usage across runs to spot inconsistent or flaky behavior.
- Description optimization: Improves the SKILL.md description so Claude invokes the skill at the right times.
- Test-case growth strategy: Encourages expanding the test set and retrying at larger scale for robustness.
Quick Start
Use this skill to create a new skill by telling it what you want Claude to do, when it should trigger, what output format you expect, and a few example prompts you consider success.
