Stop burning tokens
on a legacy design workflow

An open source Claude Code plugin that reads your brief, builds frames in Figma, implements in React, and audits the output - with you approving at every step.

/plugin marketplace add aadidesigns/agentic-design

Paste into Claude Code, restart, then run @design-planner

HOW IT WORKS

Drop in your requirements

Point it at a requirements.md and the planner agent scans for layout patterns, component inventory, and edge cases. It produces a structure brief you review before anything gets built. No assumptions, no guesswork.

$ @design-planner I have a new requirement
 
Reading inputs/requirements.md…
Inferring layout: sidebar + main content
Inventorying components: 12 found
Generating structure brief…
✓ outputs/briefs/structure.html - awaiting your approval

Build in Figma or go straight to code

Pick your path per task. The design-engineer skill talks to Figma MCP to create frames directly in your file - you review them like any other handoff. Or skip Figma entirely and let the code-engineer build React/TypeScript components from the approved spec. Either way, it runs in your local dev server.

$ @design-planner build task-03 --figma-first
 
design-engineer → creating frame in Figma via MCP
Applying design specs: typography, spacing, color palette
Frame ready - waiting for your Figma review…
✓ Approved
code-engineer → reading approved frame
Implementing React component + interactive states
✓ outputs/builds/InsightCard.tsx - live at localhost:3000

Audit only what changed

The QA skill runs visual diff, WCAG accessibility checks, and edge case coverage against the spec. Each issue comes with a token cost estimate so you decide what's worth fixing. Select the ones that matter - it generates scoped patches for just those.

$ @design-planner audit task-03
 
design-qa → checking visual consistency
Checking accessibility (WCAG AA)…
Checking edge cases: empty, overflow, loading
Mapping coverage against requirements…
✓ 2 issues found - outputs/qa-logs/task-03.md
#1 Contrast 3.8:1 → needs 4.5:1~120 tokens
#2 Missing loading state~80 tokens
Select issues to fix ▸