The TypeScript toolkit for
Agent Skills
Build, validate, lint, and ship skills that work across Claude Code, GitHub Copilot, Cursor, OpenAI Codex, Gemini CLI, and 33+ more agents.
Why Agentic Skills Framework?
The Agent Skills specification defines a portable format for giving AI agents new capabilities. It's adopted by 33+ products. This framework provides the TypeScript tooling.
Validate
Check skills against the Agent Skills specification. Catch errors before they reach production.
Lint
Go beyond spec compliance. Check context budgets, description quality, generic instructions, and progressive disclosure.
Scaffold
Generate new skills from templates. Choose from basic, with-scripts, or with-references layouts.
Get started
Install the CLI
# Install globally
npm install -g @skillscraft/cli
# Or use npx
npx @skillscraft/cli init my-skill
Use the SDK
# Install the core library
npm install @skillscraft/core
# TypeScript types
npm install @skillscraft/spec
Quick start
# Create a new skill
skill init my-skill
# Validate against the Agent Skills specification
skill validate my-skill
# Lint for best practices (context budget, description quality, etc.)
skill lint my-skill --fix
Programmatic API
import { parseSkill, validateSkill, lintSkill } from "@skillscraft/core";
// Parse a SKILL.md file
const skill = await parseSkill("./my-skill/SKILL.md");
// Validate against spec
const result = validateSkill(skill);
if (!result.valid) console.log(result.errors);
// Lint for best practices
const lint = lintSkill(skill);
lint.diagnostics.forEach(d =>
console.log(`[${d.severity}] ${d.rule}: ${d.message}`)
);
Try it in your browser
Paste a SKILL.md below to validate and lint it. Runs entirely client-side — no data leaves your browser.
Lint rules
Best-practice checks that go beyond spec compliance.
| Rule | Severity | What it checks |
|---|---|---|
context-budget | warn | SKILL.md body under 5,000 tokens / 500 lines |
description-quality | warn | Description includes "Use when" trigger clause |
no-generic-instructions | warn | Flags vague phrases like "handle errors appropriately" |
progressive-disclosure | warn | Large skills use references/ for detail |
defaults-over-menus | warn | Clear default tool/approach, not equal options |
gotchas-present | info | Suggests gotchas section for complex skills |
Resources
Concepts Guide
Deep dive into Agent Skills: progressive disclosure, activation lifecycle, cross-agent compatibility, security, and evaluation.
→Specification Reference
Complete reference for frontmatter, validation rules, lint rules, and extended spec.
→Marketplace
Browse and install production skills from SkillsCraft Hub.
→Contributing
How to contribute code, lint rules, and skill templates.
→