BACK TO SKILLS
code-qualityclaude-codecursorclinewindsurf
Skill Creator
Build, benchmark, and iteratively refine new AI agent skills using structured test suites and evaluation loops.
Skill Creator
Guide the full lifecycle of creating, testing, benchmarking, and optimizing AI agent skills.
Core Workflow
- Intent & Specification: Define what the skill should accomplish and detail its input parameters, expected outputs, and constraints.
- Drafting SKILL.md: Structure the instructions, quick-start section, and modular reference files adhering to the SKILL.md standard.
- Test Case Generation: Write parallel evaluation prompts to test performance with and without the skill enabled.
- Benchmarking: Measure pass rates, latency, and token consumption to quantify skill effectiveness.
- Trigger Optimization: Run trigger and non-trigger queries to tune frontmatter descriptions for accurate skill selection.
Installation
Add this skill to your workspace:
npx skills add https://github.com/anthropics/skills --skill skill-creator
Usage
Request skill creation or refinement assistance from your agent:
Help me create a new skill for database schema migrations and write evaluation test cases for it.