design-for-ai — Claude Code skill: building HTML interface prototypes
design-for-ai adds a four-stage design workflow to your agent: research, plan, mock, and build. Prevents generic AI aesthetics like dark cyan palettes and uniform cards.
MIT 206 installs ★ 35 updated 2026-10-21
npx skills add podo/design-agent-skills --skill design-for-ai -a claude-codeWhat design-for-ai does
The skill structures interface work into four distinct steps. First, research formalises the brief. Next, plan maps out phases, mock delivers a self-contained HTML prototype for sign-off, and build executes code phase by phase with critique steps. A seeded dealer assigns palettes and composition from reference styles, while dual-blind review flags typical AI styling tells.
The package includes doctrines covering usability heuristics and data visualisation rules. Running the dealer and detector scripts requires Node.js. Installing this package will overwrite prototype if you already use emilkowalski-skill.
What it helps with
- structure UI work through a phased four-stage process
- turn loose design ideas into a structured brief
- render self-contained HTML wireframes from design tokens
- audit interfaces against usability heuristics and chart rules
Example prompts for the agent
- Design an analytics dashboard following the research, plan, mock, and build workflow.
- Turn this rough product idea into a structured design brief.
- Create a self-contained HTML wireframe using our design tokens.
How to install design-for-ai
1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill design-for-ai -a claude-code2Manually: put the files in place
~/.claude/skills/design-for-ai/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/design-for-ai/
1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill design-for-ai -a codex2Manually: put the files in place
.agents/skills/design-for-ai/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill design-for-ai -a cursor2Manually: put the files in place
.cursor/skills/design-for-ai/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install podo/design-agent-skills skills/design-for-ai2Manually: put the files in place
.github/skills/design-for-ai/SKILL.md1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill design-for-ai -a gemini-cli2Platform CLI
gemini skills install https://github.com/podo/design-agent-skills --path skills/design-for-aiManually: put the files in place
.devin/skills/design-for-ai/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill design-for-ai -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about design-for-ai
How to install design-for-ai in Claude Code?
Run npx skills add podo/design-agent-skills --skill design-for-ai -a claude-code. It creates the skill file at ~/.claude/skills/design-for-ai/SKILL.md.
What does design-for-ai do?
It provides design vocabulary and a structured pipeline from brief to final UI. The agent moves through research, planning, interactive HTML mocking, and phased implementation without resorting to cliché AI visuals.
What prerequisites does it have?
Node.js is required to execute the palette dealer and AI-tell detector scripts. A browser MCP server is optional if you want the agent to capture screenshots of prototypes.
Source text
SKILL.md · MIT
--- name: design-for-ai description: | Design-foundations system for Claude Code built on Design for Hackers: a four-stage workflow (research, plan, mock, build) with a seeded design-DNA dealer, dual-blind review, and doctrine skills for usability, data visualisation and prototyping. triggers: - "design for AI" - "rtd design" - "design-for-ai" - "design research plan mock build" das: type: package category: design-systems upstream: "https://github.com/ryanthedev/design-for-ai" version: latest install: false --- # design-for-ai > Catalogue stub — full package: [ryanthedev/design-for-ai](https://github.com/ryanthedev/design-for-ai) <!-- das:install:start — generated from stub.yaml by `npm run generate`; edit stub.yaml, not this block --> ## Decision tree 1. **Is the full skill already installed?** This package installs 4 skills (`prototype`, `clarify`, `usability`, `data-viz`); checking the first. - Global: `f=~/.agents/skills/prototype/SKILL.md; [ -f "$f" ] && ! grep -q "^das:" "$f" && echo "installed" || echo "pointer"` - Project: `f=.agents/skills/prototype/SKILL.md; [ -f "$f" ] && ! grep -q "^das:" "$f" && echo "installed" || echo "pointer"` - `installed` → the skill is installed as `prototype`, invoke it and proceed - `pointer` → go to step 2 2. **Detect scope, then install:** ```bash [ -e ~/.agents/skills/design-for-ai ] && echo "global" || echo "project" ``` **Global:** ```bash npx skills add ryanthedev/design-for-ai --skill prototype clarify usability data-viz -g -y ``` **Project:** ```bash npx skills add ryanthedev/design-for-ai --skill prototype clarify usability data-viz -y ``` > **Claude Code:** send a command as a chat message starting with `!` to run it without leaving the conversation. **Claude Code plugin route:** ``` /plugin marketplace add ryanthedev/rtd-claude-inn /plugin install design-for-ai@rtd ``` **Heads-up** - **Requires:** Node.js for the plugin's palette, dealer and detector scripts (scripts/*.mjs); Optional: a browser MCP so the prototype skill can render and screenshot mocks - **License:** MIT per plugin.json and README; no LICENSE file in the repository - **Name collision:** `prototype` is also written by `emilkowalski-skill` (its install or its catalogue pointer); installing both overwrites the earlier one. <!-- das:install:end --> ## Invoke after install - Claude Code plugin route (the full system): `/design-for-ai:research`, then `/design-for-ai:plan`, `/design-for-ai:mock`, `/design-for-ai:build` - Skill name: `prototype` — renders a self-contained HTML mock or wireframe from design tokens and page specs - Skill name: `clarify` — turns an underspecified design request into a confirmed brief - Skill name: `usability` — heuristic evaluation, UX laws, operability questions - Skill name: `data-viz` — chart type, truthful encoding, dashboard layout - Trigger phrases: "design for AI", "rtd design", "design research plan mock build" ## What it does Teaches the agent design vocabulary instead of default AI-looking output (cyan-on-dark heroes, uniform card grids, glassmorphism everywhere). The workflow takes any design idea from brief to a viewable, gated artifact: `research` facilitates the brief, `plan` turns it into a phased design plan, `mock` renders a throwaway prototype and gates on your sign-off, and `build` executes the plan phase by phase with build, review and commit steps. A seeded dealer picks composition, hue and signature from named references (any axis can be pinned by you), and review is dual-blind: an isolated critique plus a deterministic AI-tell detector. The commands, agents and doctrine files come only with the Claude Code plugin route. `npx skills add` installs the four standalone skills; `usability` and `data-viz` are doctrine files the workflow reads, so on Claude Code they are marked as not directly invocable. ## When NOT to use - Lightweight UI polish on an existing component — use `taste-skill` for faster, lower-overhead iteration - Accessibility-specific audits focused on ARIA, keyboard navigation, or WCAG compliance — use `fixing-accessibility` or `wcag-audit-patterns`
© podo · MIT · podo/design-agent-skills