Draft SKILL.md files and refine them through prompt-based test evaluations.
npx skills add https://github.com/zhboner/Agent-Skills --skill skill-creator-zhboner
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/zhboner/Agent-Skills/tree/main/Global-skills/skill-creator Command: npx skills add https://github.com/zhboner/Agent-Skills --skill skill-creator-zhboner
This Skill helps you go from a rough idea to a working agent Skill by drafting, testing with realistic prompts, and iterating based on quantitative benchmarks and qualitative review.
Tell the skill where you are in the process (e.g., “I want to create a skill for X” or “I already have a draft and want to improve it”), then answer the initial interview questions about triggers, output format, and evaluation preferences.
High-intent search queries and answers about installing and using this skill.
You can create and test agent workflows by drafting a SKILL.md from user intent, running prompt-based test cases with and without the skill, and refining instructions using qualitative outputs and quantitative metrics.
Skill benchmarking evaluates agent performance by running verifiable eval expectations to analyze pass rate, timing, and resource usage, which helps identify regressions and gains across test cases.
To optimize trigger descriptions, generate trigger eval queries and iteratively improve the SKILL.md frontmatter description so the agent invokes the skill at the correct moments based on user intent.
Yes, you can run prompt-based test cases with and without the skill to compare behavior, using benchmarking aggregation and qualitative review to identify regressions and gains.
The best way to refine skill instructions is to run test cases, collect quantitative metrics and qualitative outputs via benchmarking aggregation, and iteratively update workflow instructions and trigger descriptions based on eval results.
You can start skill development from a rough idea without an existing draft by answering interview questions about triggers, output format, and evaluation preferences to generate a draft SKILL.md.
8b-isDesign and refine Claude Code Skills with consistent SKILL.md structure.
C0ntr0lledCha0sImprove Claude Code component descriptions, permissions, triggers, and documentation.
derekhu0002Automates eval-driven development for Claude Code sessions with pass@k metrics and regression suites.
adriannoesRun iterative mutations and binary scoring to refine Claude Code skill prompts.
