SkillCodex is a curated skills system for AI agents: structured SKILL.md instructions that turn vague prompts into repeatable workflows with clear outcomes, safety boundaries, and modern dev tooling support. Built for Cursor, Claude Code, skills.sh, and any host that reads agent skills.
Use it as a developer productivity tool and lightweight automation framework for frontend teams - not a black-box plugin. Every skill is readable markdown on GitHub, with an optional npm package for apps and CI.
███████╗██╗ ██╗██╗██╗ ██╗ ███████╗
██╔════╝██║ ██╔╝██║██║ ██║ ██╔════╝
███████╗█████╔╝ ██║██║ ██║ ███████╗
╚════██║██╔═██╗ ██║██║ ██║ ╚════██║
███████║██║ ██╗██║███████╗███████╗███████║
╚══════╝╚═╝ ╚═╝╚═╝╚══════╝╚══════╝╚══════╝
Install into Cursor (.cursor/skills/), Claude Code, or skills.sh-compatible hosts.
npm install @skillcodex/skills
# or
pnpm add @skillcodex/skillsRequires Node.js 18+.
git clone https://github.com/bh611627/skillcodex.git
cd skillcodex
pnpm install
pnpm validate- Pick a skill from the directory or table below.
- Run
npx skills add … --skill <slug>or copyskills/<slug>/SKILL.mdinto your agent skills folder. - Invoke with a concrete task (see examples).
import seoAgent from "@skillcodex/skills/seo-agent";
seoAgent.name; // "seo-agent"
seoAgent.description; // one-line summary
seoAgent.instructions; // agent body
seoAgent.outcomes; // measurable results
seoAgent.skillMd; // full SKILL.md sourceimport { skills, skillList } from "@skillcodex/skills";More detail: docs/usage.md
| Goal | Skill | Example prompt |
|---|---|---|
| Next.js SEO + metadata | seo-agent | “Audit my About page metadata and add JSON-LD” |
| Small focused code diffs | code-assistant | “Fix login button styles without rewriting the file” |
| Vitest + RTL tests | testing-agent | “Add tests for UserCard.tsx” |
| WCAG 2.2 AA audit | accessibility-audit | “File:line a11y report for the checkout form” |
| JS → strict TypeScript | typescript-refactor | “Migrate utils/ to strict TS” |
| Core Web Vitals / bundle | performance-audit | “Find LCP and bundle issues in my App Router app” |
| Safe env setup | env-config-agent | “Create .env.example and typed env with zod” |
| Prisma schema | database-schema-agent | “Review schema for missing indexes” |
| Doc UI (skills.sh-style) | web-design-guidelines | “Build a skill browser with skeletons and outcomes” |
Step-by-step host guides: Cursor · Claude Code
Without skill: vague advice, full-file rewrites, mixed npm/pnpm
With skill: scoped diff, file:line findings, stated outcomes, safety limits
- Reusable workflows - capture how your team actually ships (SEO, a11y, TS strictness) once, reuse everywhere.
- Trust - full
SKILL.mdon GitHub; validation catches prompt-injection Unicode and missing safety sections. - Modern dev tooling - pnpm, Next.js App Router, Vitest, Playwright, Prisma patterns baked in.
- AI + human readable - structured for agents and code review in PRs.
Compared to skills.sh: same npx skills add discovery, but outcomes instead of install counts, and every skill is auditable markdown.
| Skill | Outcome |
|---|---|
| web-design-guidelines | Doc UI build or file:line design audit |
| skillcodex-browser-ui | Full documentation browser UI |
| content-creator | Hooks, script, caption, hashtags |
| seo-agent | SEO outline, JSON-LD, OpenGraph checklist |
| code-assistant | Focused diffs + code review |
| skill-creator | Valid new SKILL.md + metadata |
| typescript-refactor | Strict TS migration or audit |
| testing-agent | Vitest/RTL + Playwright tests |
| accessibility-audit | WCAG 2.2 AA report |
| env-config-agent | Safe env + typed config |
| performance-audit | Bundle + CWV findings |
| database-schema-agent | Prisma schema or audit |
| Path | Purpose |
|---|---|
skills/ |
Source of truth - one SKILL.md per skill |
references/ |
Shared rules (design, SEO, safety) |
package/ |
npm package @skillcodex/skills |
docs/ |
Introduction, usage, contributing |
examples/ |
Cursor and Claude Code guides |
REPO_LAYOUT.md · GIT_WORKFLOW.md · CHANGELOG.md
| Doc | Description |
|---|---|
| docs/introduction.md | Concepts and ecosystem |
| docs/usage.md | Install paths and npm API |
| docs/contributing.md | Add or improve skills |
| SKILL_STANDARD.md | SKILL.md format spec |
| WHAT_IS_SKILLCODEX.md | Short overview |
| SECURITY.md · TRUST.md | Safety model |
- api-route-agent - Route Handlers + zod validation
- i18n-agent - App Router i18n setup and audit
- auth-agent - Session/OAuth without secret leakage
- monorepo-agent - pnpm workspaces + Turborepo
- storybook-agent - Stories aligned with design-guidelines
PRs welcome. See .github/CONTRIBUTING.md and run pnpm validate before opening a PR.
pnpm new-skill # interactive scaffoldMIT © SkillCodex contributors