Create and iteratively improve Claude Code skills with test-driven evaluation.
npx skills add https://github.com/iadr-dev/colab --skill skill-creator-iadr-dev
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/iadr-dev/colab/tree/main/skills/skill-creator Command: npx skills add https://github.com/iadr-dev/colab --skill skill-creator-iadr-dev
This Skill helps you create new Claude Code skills, iteratively improve existing ones, and objectively measure whether a skill triggers correctly for the right user intents.
Use the skill creator to build a new skill by answering the questions about what the skill should do, when it should trigger, and what output format you want, then generate and run a small test set to start the eval/iterate loop.
High-intent search queries and answers about installing and using this skill.
Create Claude Code skills by drafting instructions, generating test prompt sets, running trigger evaluations, and grading expectations to benchmark performance. Use the eval viewer to review aggregated statistics and iterate until triggering accuracy improves.
Benchmark Claude skill triggering accuracy by generating test prompts, running Claude with skill access, and comparing behavior against expected triggers. Variance-aware reporting aggregates statistics to identify under-triggering and false triggers for description optimization.
Optimize SKILL.md descriptions by running trigger evaluations against test prompt sets, reviewing quantitative results in the eval viewer, and iteratively refining the description. This reduces both under-triggering and false triggers for target user intents.
Modify and benchmark existing Claude skills by generating test prompts, running trigger evaluations, and comparing behavior against expected outcomes. The iterative workflow supports refining draft instructions and descriptions for skills you already have.
The test harness runs Claude-with-access-to-the-skill against generated eval sets, grades expectations, and aggregates benchmark statistics with variance-aware reporting. Results are reviewed in the eval viewer to drive repeat iterations.
Prompt engineering experience helps but is not required. The skill creator guides you through answering what the skill should do, when it triggers, and output format, then generates and runs test sets to start the eval and iterate loop.
krzemienskiOrchestrate Claude Code skill development from SKILL.md drafting to trigger evaluation.
oguzhanguvenkayaOrchestrate end-to-end creation and optimization of Claude skills with evaluation loops and benchmarking.
dyCuong03Draft, test, grade, and optimize Claude Code skills with benchmark evaluations.
zhbonerDraft SKILL.md files and refine them through prompt-based test evaluations.
