📦 124 production-ready AI agent skills with model-specific variants for Claude, GPT, GLM, and DeepSeek.
Build better AI agents using reusable, documented, and model-optimized skills.
This repository is framework-agnostic. The skills can be used with Claude Code, OpenAI Codex, Cline, Cursor, Roo Code, Continue, OpenHands, or custom AI agent systems.
- 📦 124 organized agent skills
- 🗂️ 10 practical categories
- 🤖 620 model-specific variants across 5 model families
- 📖
README.md+SKILL.mdfor every canonical skill - 🧾 Manifest and migration map included
- 🔌 Framework-agnostic: copy, adapt, or integrate anywhere
- Browse
organized-skills/. - Choose a category.
- Open a skill folder.
- Read
README.mdfor the human-friendly explanation. - Use
SKILL.mdas the actual agent instruction.
organized-skills/
↓
engineering-workflows/
↓
your-skill/
├── README.md # explains the skill
└── SKILL.md # use this in your agent
If you already know your target model, start from model-specific-skills/ instead.
.
├── organized-skills/ # Canonical source of truth
│ ├── SKILLS.md # Human-readable skill index
│ ├── skills-manifest.json # Machine-readable manifest
│ ├── MIGRATION_MAP.md # Old path → new path mapping
│ └── <10 category folders>/
│
├── model-specific-skills/ # Model-adapted skill versions
│ ├── claude/
│ ├── gpt-5-5/
│ ├── gpt-5-4-and-earlier/
│ ├── glm/
│ └── deepseek/
│
├── SKILLS_OVERVIEW.md # Full detailed overview
└── FINAL_CLEANUP_REPORT.json # Cleanup verification summary
| Category | Skills | What it covers |
|---|---|---|
frontend-ui-design |
24 | UI design, styling, design systems, themes, accessibility, visual audits |
responsive-mobile-navigation |
12 | Mobile menus, drawers, touch behavior, viewport safety, responsive UI |
website-business-builds |
6 | Website rebuilds, business sites, localized pages, static galleries |
brand-content-creation |
12 | Brand identity, banners, slides, articles, copy polish, content workflows |
engineering-workflows |
15 | Implementation, architecture, domain modeling, prototypes, handoffs |
code-quality-testing |
16 | Debugging, tests, reviews, refactors, merge conflicts, git guardrails |
product-planning-issues |
10 | PRDs, issues, triage, QA, decision maps, planning workflows |
knowledge-learning-writing |
13 | Teaching, notes, writing, glossary, learning and knowledge workflows |
automation-integrations-tools |
3 | Setup helpers, integrations, plugin/tool workflows |
skill-authoring-meta |
13 | Writing, routing, maintaining, and improving skill systems |
The canonical skills live in organized-skills/. The same skills are also rewritten for different model styles:
| Model folder | Style |
|---|---|
claude |
Structured, reasoning-friendly, clear constraints |
gpt-5-5 |
Compact, direct, high-signal instructions |
gpt-5-4-and-earlier |
More explicit, step-by-step, lower ambiguity |
glm |
Short, concrete, numbered, simple do/do-not rules |
deepseek |
Detailed procedure, validation, failure handling |
Model-specific versions preserve the same skill behavior. Only the writing style and instruction structure are adapted.
Want an agent to debug a hard bug?
organized-skills/
↓
code-quality-testing/
↓
diagnosing-bugs/
├── README.md
└── SKILL.md
Using Claude?
model-specific-skills/
↓
claude/
↓
code-quality-testing/
↓
diagnosing-bugs.md
SKILLS_OVERVIEW.md— full detailed overvieworganized-skills/SKILLS.md— complete skill indexorganized-skills/skills-manifest.json— machine-readable manifestorganized-skills/MIGRATION_MAP.md— migration maporganized-skills/migration-map.json— machine-readable migration mapFINAL_CLEANUP_REPORT.json— cleanup summary
Organized skills: 124
Categories: 10
Model-specific sets: 5
Model-specific skill files: 620
Needs review after cleanup: 0
Think of this repository like a cookbook for AI agents:
- A skill is a recipe.
- A category is a cookbook chapter.
- A model-specific version is the same recipe rewritten for a different AI model.
Start with organized-skills/SKILLS.md, then open the category that matches your task.
License not specified yet. Add a LICENSE file before public release if needed.
