An agentic skills framework & software development methodology that works.
UI and UX skill pack for Claude Code and Codex
OJO Design Skills gives an agent a reusable UI/UX playbook for design-led output. It uses a two-track method: one path for established product patterns and one for brand-driven design, then turns the chosen direction into tokens, components, motion, and final specs.
Builders who want their agent to make clearer design choices and avoid generic interface output.
You can get more intentional UI and UX specs from your agent instead of re-explaining design direction every time.
What it does
Two-track methodology
Chooses between a Convention Track for proven SaaS patterns and an Innovation Track for brand-driven identity work.
Anti-slop guardrails
Blocks common weak patterns like purple-blue gradients, gray-box placeholders, and single-hue palettes.
Real imagery requirement
Requires real subject-specific imagery when a screen uses images, instead of fake stock photos or decorative blobs.
Code-grade design output
Produces atomic Tailwind class strings, an 8-state interaction model, and spring-based motion specs.
Single installer for multiple clients
Uses one install script to place the skill bundle into Codex, Claude Code, ZCode, and other supported clients.
Mandatory style-direction gate
Asks for 2–3 distinct design directions before generating tokens and specs.
How to get it
- 1Install all skills with one command — replace with your agent
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target <client>
- 2Replace existing copies without backups
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target codex --force
- 3Install to a custom client home
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | CLAUDE_HOME=/path/to/.claude bash -s -- --target claude-code
- 4Install to an explicit directory
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target opencode --dest /path/to/skills
- 5Install from a local checkout (useful for development)
./scripts/install.sh
- 6Install from a different local checkout
./scripts/install.sh --source /path/to/OJO-Design-Skills --target codex
README
OJO Design Skills
Reusable UI/UX design skills for AI coding agents — Codex, Claude Code, ZCode, and more.
OJO Design Skills is a drop-in skill bundle that teaches your AI coding agent how to make real design decisions — not generic "AI-looking" layouts. Install once, and your agent gains a complete UI/UX methodology: brand-driven visual direction, design tokens, component recipes, motion specs, and a strict anti-pattern guardrail that keeps output from looking like slop.
It is agent-agnostic and ships with a single installer that targets every major client.
Features
- Two-track methodology. Convention Track adopts a proven design language for SaaS and utility products; Innovation Track builds brand-driven identity via Material Metaphor, archetype, narrative, or cultural-semiotic approaches.
- Anti-AI-slop guardrails. Hard-banned combinations (purple-blue gradients, gray-box placeholders, single-hue palettes) are enforced as rules, not suggestions.
- Real-imagery requirement. Any screen that shows imagery must use real, subject-specific assets — no decorative placeholders, fake stock smiles, or gradient blobs.
- Code-grade output. Components resolve to atomic Tailwind class strings with an 8-state interaction model; motion uses spring physics, not fixed curves.
Showcase
The same brief, with and without these skills installed.
| With OJO Skills | Without OJO Skills |
|---|---|
![]() | ![]() |
![]() | |
Design samples
Additional output produced with these skills installed — one product, multiple surfaces.




Quick start
Install all skills with one command — replace <client> with your agent:
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target <client>
The remote installer requires Bash, curl, and tar.
| Client | --target | Default install directory |
|---|---|---|
| Codex | codex | ${CODEX_HOME:-~/.codex}/skills |
| Claude Code | claude-code | ${CLAUDE_HOME:-~/.claude}/skills |
| ZCode | zcode | ${AGENTS_HOME:-~/.agents}/skills |
| DeepCode | deepcode | ${AGENTS_HOME:-~/.agents}/skills |
| WorkBuddy | workbuddy | ${WORKBUDDY_HOME:-~/.workbuddy}/skills |
| OpenCode | opencode | ${OPENCODE_CONFIG_DIR:-~/.config/opencode}/skills |
| Generic agent | generic | ${AGENTS_HOME:-~/.agents}/skills |
Restart or reload the client if it does not pick up the new skills.
Advanced install options
Replace existing copies without backups:
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target codex --force
Install to a custom client home:
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | CLAUDE_HOME=/path/to/.claude bash -s -- --target claude-code
Install to an explicit directory:
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target opencode --dest /path/to/skills
Install from a local checkout (useful for development):
./scripts/install.sh
Install from a different local checkout:
./scripts/install.sh --source /path/to/OJO-Design-Skills --target codex
Install a pinned tag or commit for a reproducible remote install:
curl -fsSL https://raw.githubusercontent.com/touchine-ojo/OJO-Design-Skills/main/scripts/install.sh | bash -s -- --target codex --ref <tag-or-commit>
Preview what would be written without changing files:
./scripts/install.sh --target opencode --dry-run
Install a single skill (--skill can be repeated):
./scripts/install.sh --target zcode --skill app-ui-ux-best-practices
Skills
app-ui-ux-best-practices
The core methodology. Picks a track based on the product's value proposition, then produces complete design specifications.
- Convention Track — adopt one proven design language (Notion, Linear, Stripe, Things 3, …) for SaaS and utility products.
- Innovation Track — build a brand-driven identity via Material Metaphor, archetype, narrative, or cultural-semiotic methods.
- Ships with 9 reference files: anti-patterns, visual tokens, component recipes, motion system, material metaphor, design audit, and more.
How it works
Product brief
│
├─ Convention Track ──► pick proven system ──► tokens ──► components ──► specs
│ (SaaS / utility) (Notion, Linear…)
│
└─ Innovation Track ──► insight + feeling ──► methodology ──► tokens ──► specs
(brand-driven) (Material Metaphor / archetype / narrative)
Both tracks require a mandatory style-direction confirmation gate before producing tokens — the agent presents 2–3 genuinely distinct directions and waits for selection. Once installed, skills are activated automatically when the agent detects design-related intent (e.g. "design system", "视觉设计", "design tokens").
Contributing
Skills are plain Markdown files under skills/<name>/SKILL.md plus optional references/ files. To add or refine a skill, read the contributing guide, edit the relevant files, and run the installer locally to verify:
./scripts/install.sh --dry-run
Use GitHub Discussions for questions and early ideas, and Issues for reproducible bugs or focused feature requests. Security concerns should follow the private process in SECURITY.md.
License
Files in the repo
- .github
- docs
- scripts
- skills
- .gitignore
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- LICENSE
- README.md
- SECURITY.md
- SUPPORT.md
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More skills

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. Chinese or English; every frame drawn in code with Remotion.
Public repository for Agent Skills
Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)

Production-grade engineering skills for AI coding agents.


