Skill detail
product-design
Useful for Figma-based handoff and design-system implementation planning, but setup-dependent.
Inspect before use
Automated review checks relevance, not safety or endorsement. Read the source instructions before using this skill.
SKILL.md
The saved excerpt is a snapshot from review. The external source remains the complete and most current version.
---
name: product-design
description: Automates design review, token extraction, component mapping, and implementation planning. Reduces design handoff from 6-10 hours to 5 minutes via direct Figma MCP integration. Auto-invoke when user mentions design review, Figma mockup, or design handoff.
allowed-tools: Read, Write, Edit, Grep, Glob, Bash, Task, TodoWrite
version: 1.1.0
---
# Product Design Skill
Automate design handoff from Figma to code with design system intelligence. Extract tokens, map components, detect drift, generate implementation plans.
## When to Invoke
Auto-invoke when user says:
- "Review this design"
- "Analyze Figma mockup"
- "Design handoff for [feature]"
- "Check design system impact"
- "Plan implementation for design"
- "Extract tokens from Figma"
- "What changed in the design?"
## What This Does
**5-Step Workflow**:
1. **Design Analysis**: Extract patterns, components, tokens from Figma
2. **Codebase Audit**: Compare design vs implementation, find drift
3. **Implementation Planning**: Generate phased task breakdown
4. **Task Assignment**: Create Navigator task document
5. **Handoff**: Ask user to review or start implementation
**Time Savings**: 6-10 hours → 15-20 minutes (95% reduction)
## Prerequisites
### Required
1. **Python Dependencies**
```bash
cd skills/product-design
./setup.sh # Automated installation
# OR manually: pip install -r requirements.txt
```
2. **Figma Desktop** (for automated workflow)
- Download: https://www.figma.com/downloads/
- Enable MCP: Figma → Preferences → Enable local MCP Server
- Must be running during design reviews
3. **Project Structure**
- `.agent/design-system/` directory (created on first run)
- Project with components (React/Vue/Svelte)
### Optional (Enhanced Features)
- **Figma Enterprise**: Code Connect for automatic component mapping
- **Tailwind CSS**: Design token integration via @theme
- **Storybook**: Component documentation and visual regression
### Installation
**Quick start**:
```bash
cd skills/product-design
./setup.sh
```
See `INSTALL.md` for detailed installation guide and troubleshooting.
## Workflow Protocol
### Step 0: Check Setup (Auto-Run)
**Before starting, verify Python dependencies installed**:
```bash
# Get Navigator plugin path
PLUGIN_PATH=$(dirname "$(dirname "$(dirname "$PWD")")")
# Check if venv exists
if [ ! -d "$PLUGIN_PATH/skills/product-design/venv" ]; then
echo "❌ product-design skill not set up"
echo ""
echo "Run setup (30 seconds):"
echo " cd $PLUGIN_PATH/skills/product-design && ./setup.sh"
echo ""
echo "Or use manual workflow (no Python needed)"
exit 1
fi
```
**If setup missing**:
- Show setup instructions
- Offer manual workflow as alternative
- **Do not proceed** with automated Figma workflow
**If setup complete**:
- Continue to Step 1 (Design Analysis)
---
### Step 1: Design Analysis
**Objective**: Extract design patterns from Figma or manual description
#### With Figma MCP (Automated) ✨ SIMPLIFIED
**New Architecture** (v1.1.0+): Python directly connects to Figma MCP - no manual orchestration!
```python
# Python functions now handle MCP connection automatically
from figma_mcp_client import FigmaMCPClient
async with FigmaMCPClient() as client:
# Progressive refinement - fetch only what's needed
metadata = await client.get_metadata()
components = extract_components(metadata)
# Fetch details only for complex components
for comp in components:
if comp['complexity'] == 'high':
comp['detail'] = await client.get_design_context(comp['id'])
# Get design tokens
variables = await client.get_variable_defs()
```
**Workflow** (fully automated):
1. User provides Figma URL
2. Run `python3 functions/design_analyzer.py --figma-url <URL>`
3. Python connects to Figma MCP (http://127.0.0.1:3845/mcp)
4. Fetches metadata → analyzes → fetches details only if needed
5. Returns complete analysis
**Benefits**:
- ✅ No manual MCP toRead the full source on GitHub (opens external page)