New Skill Scaffolder
Create a new Claude Code skill correctly: right mechanism, valid name and folder, a triggering description, and a smoke test.
Before & after
Same prompt, same agent — the only difference is the skill.
Setting up a skills repo is pretty straightforward but there are a lot of little decisions to make. You'll want to define the structure, set up documentation templates, create guidelines for how skills are documented, decide on version control strategy, and maybe set up some automation. It depends on what kind of skills you're tracking and who needs access. Want me to walk you through a basic setup?
SKILLS REPOSITORY INITIALIZATION
REPO STRUCTURE
.
├── README.md (overview + quick start)
├── CONTRIBUTING.md (contribution workflow)
├── skills/
│ ├── web-frontend/
│ │ ├── SKILL.md (React, TypeScript, testing)
│ │ ├── resources/
│ │ │ ├── interview-questions.md
│ │ │ └── reading-list.md
│ │ └── examples/
│ │ ├── component-example.tsx
│ │ └── test-example.spec.ts
│ ├── backend/
│ ├── devops/
│ └── product/
├── templates/
│ └── SKILL-TEMPLATE.md
└── .github/
└── workflows/
└── validate-skills.yml
DOCUMENTATION TEMPLATE (per skill)
• Title and owner
• Proficiency levels (tier 1-3 rubric)
• Learning outcomes
• Recommended resources (books, courses, projects)
• Assessment questions
• Hands-on project examples
• Time to proficiency estimate
AUTOMATION
• GitHub Actions: validate SKILL.md format on PR
• Linter checks: required sections present
• Auto-generate index.md across all skills
ACCESS & PERMISSIONS
• Public repository (read-only for external)
• Team write access via GitHub organization
• Annual review cycle for content updates
INITIAL POPULATION
Staged skill uploads: 8 core skills in week 1, expanded quarterly
READY FOR: Distributed onboarding, skill assessments, hiring alignmentAbout this skill
Most new skills fail because they should have been a CLAUDE.md line or a hook, have a description that never triggers, or restate steps the model already follows. This scaffolder designs against all three and gives you the complete SKILL.md, writing files only when you ask.
What it does
- Recommends the right mechanism first: skill, CLAUDE.md, hook, subagent, or user-invoked command
- Picks location and a valid name that matches its folder, and checks for collisions
- Drafts a trigger-ready description with real user phrasings
- Builds the body around domain knowledge, with reference files and scripts for progressive disclosure
- Includes smoke-test prompts for a fresh session: a trigger, a near-miss, and a quick question
What you get
A ready-to-save skill (SKILL.md plus any reference files or scripts), created in your repo on request, its description, and smoke-test results or commands to run.
Best for
Anyone turning a repeated workflow into a Claude Code skill, or setting up a project's first skills folder.
SKILL.md preview
---
name: skill-scaffold
description: Create a new Claude Code skill (or a project's first skills folder) the right way - first check a skill is the right tool versus CLAUDE.md, a hook, or a subagent, then pick the install location, a valid name that matches its folder, a description that will actually trigger, a SKILL.md built around domain knowledge rather than generic steps, and reference files or scripts for progressive disclosure - then smoke-test it in a fresh session. Use when the user says "make a skill for X", "turn this into a skill", "scaffold a new skill", "set up .claude/skills in this repo", "save this workflow as a skill", or "how should I structure my skill?".
---
… (sign up to view the full skill)More ai & agents skills
View all AI & Agents skills →Skill Library Lint and Overlap Report
Lint every skill in a repo at once: frontmatter, name-folder match, weak descriptions, overlapping triggers, broken refs, plus a CI script.
Self-Verifying Agent Loop: Plan, Check, Retry, Close
Turns a goal into a task plan with machine-run checks, executes with retry caps, escalates, and refuses to close until verified.
Skill Test Runner (Trigger and A/B)
Test a skill in fresh sessions: ready-to-run prompts, commands, and rubric for trigger recall, false triggers, and blind A/B.
Chief AI Officer Plan Review
Evidence-first review of one AI plan: eval gaps, cost per successful task, model and vendor risk, prompt injection, and EU AI Act screen.