Product Team Claude Skills · 2026

Claude Code for Product Designers

A Figma-powered workflow that turns Claude into a design-aware partner — Vuetify-compatible, policy-enforcing, and context-rich from day one.

9
Designer Skills
6
Shared Skills
1
Chain Command
15
Total Commands
Context

Why generic AI tools fall short for designers

Design reviews, Figma handoffs, and Vuetify constraint checks require deep product and system knowledge that a generic assistant doesn't have.

Pain point 1

No design system awareness

Generic Claude doesn't know which Vuetify components your product uses, what's been customized, or what gaps exist in the design system.

Pain point 2

No policy context

Design decisions must follow product policy (RTL, accessibility baseline, state coverage). A generic assistant can't enforce rules it doesn't know about.

Pain point 3

Disconnected from Figma

Describing a design in chat is slow and lossy. The context should come directly from the Figma file — components, tokens, flows, and states.

Pain point 4

Scope overreach

Generic Claude suggests adding scope beyond what the PM approved in the wireframe — creating tension between design and product management.

Design Principle

4-Layer Context Assembly

Same 4-layer model as the PM workflow. The key difference: Layer 2 (the interview) is Figma-powered — Claude reads the design file directly instead of asking questions.

Layer 1 · Policy
Skills, commands, rules, output style
~/.claude/ and claude-workflow/policy/
Fixed · npm update
Layer 2 · Figma Interview
Automated Figma-based onboarding
claude-workflow/interview/designer-interview.md · reads Figma via MCP
Figma-powered · run once
Layer 3 · Dynamic
CLAUDE.md + 7 context files + 2 design agents
CLAUDE.md, .claude/skills/product-designer/*/context.md
Generated · unique per designer
Layer 4 · Personal
Private notes and local settings
CLAUDE.local.md, .claude/settings.local.json · gitignored
Private · never committed

Key difference from PM: The designer interview doesn't ask questions. It connects to Figma via MCP, reads your design system and product screens directly, and generates all 11 context files from the Figma file content.

Files · What Each One Does

Every file, explained

The designer setup generates 11 files — fewer than the PM (14), because Figma data replaces several manual interview answers.

Global — ~/.claude/
SK
skills/product-designer/*/SKILL.md
8 files · design review, Figma-to-code, constraint check, microcopy, handoff, research, policy review, implementation review
SK
skills/shared/*/SKILL.md
6 files · wireframe-generator, design-system-check, task-writer, presentation-style, meeting-support, prompt-optimizer
CM
commands/product-designer/*.md
6 slash commands · each chains one or more skills
OS
output-styles/design-standard.md
Response format with flag system (🔴 Blocker, 🟡 Important, 🟢 Suggestion) and state coverage table
IV
commands/start-designer-interview.md
Entry point for designer onboarding · triggers the Figma-based interview
Project-level — inside product repo
CL
CLAUDE.md
Designer profile + product design context + working language · generated from Figma
CX
skills/*/context.md
7 files · design system details, Vuetify constraints, component inventory — extracted from Figma
AG
agents/design-agent.md
Primary design agent · handles reviews, checks, and handoffs
AG
agents/handoff-agent.md
Specialized handoff agent · ensures engineering receives complete, implementable specs
ST
settings.json
Claude Code permissions for this repo · includes Figma MCP access settings
OP
outputs/
Saved deliverables · design reviews, handoff docs, QA reports — MD or HTML
PR
CLAUDE.local.md
Private notes · gitignored · personal context loaded alongside CLAUDE.md
Request Assembly

How a Claude request is formed

The same 4-layer merge applies to every designer command. Here's what happens when you run /design-review.

1

CLAUDE.md is loaded automatically

Claude Code injects CLAUDE.md into every session. It contains the designer's profile, product design language, Vuetify version, design system boundaries, and working language.

2

Command file is read

~/.claude/commands/product-designer/design-review.md defines the chain: design-policy-review → vuetify-constraint-check → unified review report. Each step has a defined output format.

3

SKILL.md is read for each skill

~/.claude/skills/product-designer/design-policy-review/SKILL.md defines what to check and how — Vuetify coverage check, RTL layout rules, state coverage requirements, microcopy standards, accessibility baseline.

4

context.md is read for the same skill

.claude/skills/product-designer/design-policy-review/context.md contains your product's design system — component inventory extracted from Figma, known gaps, customized components, and team-specific rules.

5

Layers merge → a complete, contextualized review

SKILL.md (check instructions) + context.md (your Figma-extracted design system) + CLAUDE.md (designer profile) + design-standard.md (output format with flag system) = a review that knows your product, your components, and your rules.

SKILL.md answers "how to review". context.md answers "against which design system and constraints". The context was extracted from Figma — no manual description needed.

Layer 2 · Figma Interview

The designer interview — automated from Figma

Unlike the PM interview (647 lines of questions), the designer onboarding connects to Figma via MCP and reads the design system and product screens directly. Two questions. No manual context entry.

1

Verify Figma MCP access

Checks whoami via Figma MCP to confirm access is available before starting. If Figma MCP is not connected, provides setup instructions.

2

Two quick questions

Working language (Persian or English) and whether the designer wants to connect an existing git repository. These are the only manual inputs required.

3

Designer provides Figma file links

Design system file link + one or more product screen links. These are the source of truth for all generated context.

4

Claude reads Figma directly

Uses Figma MCP tools to read components, color tokens, text styles, spacing variables, component variants, and screen flows. No copy-pasting, no manual description.

5

11 files generated and written

CLAUDE.md, design-agent.md, handoff-agent.md, design-standard.md, and 7 skill context files. Designer approves each write. Context reflects the actual Figma file.

11
Generated files
2
Manual questions
Figma
Source of context
Run once
Layer 1 · Policy

14 skills — 8 designer · 6 shared

Shared skills are available to both the PM and the Designer. Designer-specific skills cover the full design workflow from research to handoff.

Review — design quality and policy compliance

design-policy-review

Checks design against policy: Vuetify coverage, RTL layout, state coverage, microcopy quality, and accessibility baseline.

vuetify-constraint-check

For every UI component, verifies Vuetify 3 coverage. Maps each element to a component, checks theming, flags gaps with options.

design-qa

QA-perspective review of a design: checks all states are covered, identifies implementation ambiguities before handoff.

Produce — artifacts and deliverables

design-handoff

Generates a complete handoff document for engineering with component mapping, state specs, and Vuetify prop recommendations.

figma-to-code

Translates Figma screens into Vue + Vuetify code with correct component choices, props, and responsive behavior.

microcopy-writer

Writes UI copy (labels, placeholders, error messages, empty states) following product voice and tone guidelines.

Research + Post-Implementation

design-research

Structures research questions, synthesizes user feedback, and surfaces insights in a format PMs can act on.

implementation-review

Compares Figma design against the built Vue/Vuetify component — identifies gaps and produces a structured report.

Shared (PM + Designer)
Shared

design-system-check

Scans design system before new UI is requested — prevents duplicate components.

Shared

wireframe-generator

Mid-fidelity HTML wireframe covering all states using existing design system components.

Shared

task-writer

Converts design decisions or review findings into ready-to-paste Linear tasks with four-part structure.

Shared

presentation-style

Generates polished single-file HTML slide decks following the product design system.

Shared

meeting-support

Prepares structured agendas and extracts decisions and action items from design critiques.

Shared

prompt-optimizer

Analyzes and rewrites prompts to improve Claude output quality for design reviews and requests.

Layer 1 · Policy

6 slash commands

Designer commands cover the full design cycle — from research and review through Figma-to-code translation and final handoff.

/design-review
design-policy-review → vuetify-constraint-check → unified review report with blockers, improvements, and required actions
2 skills
/design-handoff
Runs design-qa then generates a complete handoff document for engineering — component specs, state coverage, Vuetify props
2 skills
/figma-to-code
Reads a Figma screen via MCP and translates it into Vue + Vuetify 3 code following your design system conventions
1 skill
/design-qa
QA-lens review of a design before handoff — all states covered, no implementation ambiguities, no blockers outstanding
1 skill
/design-research
Structures research questions and synthesizes user feedback into PM-ready insights with confidence levels
1 skill
/microcopy-writer
Writes UI copy for a screen or component following product voice — labels, placeholders, errors, empty states, confirmations
1 skill
Designer Skill · Discovery

design-research

Understand the design context before starting visual work — existing patterns, component inventory, and open UX decisions that could affect the design direction.

Designer Skill Discovery Phase Standalone
When to use

"What components are already in use", "what design patterns exist for this", "before I start designing", or picking up a feature for the first time. Typically the first skill a designer runs on a new feature.

What it outputs

Existing design patterns relevant to the feature, component inventory (available vs. missing), open UX decisions that need to be made, and a starting checklist before running design-policy-review.

Invoke: /design-research [feature or area to research]

Designer Skill · Review · Mid-Process

design-policy-review

Check design work-in-progress against design-policy.md while changes are still cheap — before the design is finalized and rework becomes expensive.

Designer Skill Mid-Process Only Step 1 in /design-review
When to use

"Check my design direction", "is this RTL-safe", "does this follow our rules", "I'm designing this, does it look right so far". Use WHILE designing — not after. Do NOT use on a completed design; use design-qa for that.

What it outputs

Policy compliance status with severity flags (🔴 Blocker / 🟡 Important / 🟢 Suggestion). Checks: Vuetify coverage (surface), RTL layout, state coverage, microcopy quality, accessibility baseline.

Not the same as design-qa: policy-review is mid-process; design-qa is the final gate check before handoff.

Designer Skill · Compatibility

vuetify-constraint-check

Systematically verify that every UI element in a design has a corresponding Vuetify 3 component — flag any gaps before the design goes to engineering.

Designer Skill Vuetify 3 Step 1 in /design-handoff
When to use

"Check Vuetify compatibility", "can this be built with Vuetify", "verify component coverage", "are there any Vuetify gaps", or as Step 1 in /design-handoff before generating the handoff document.

What it outputs

A component coverage map: each UI element → Vuetify 3 component + key props + theming compliance. Gaps flagged with options: extend existing component, use a different component, or custom implementation needed.

Runs as Step 2 in /design-review (after policy check) and as Step 1 in /design-handoff (before handoff doc). Ensures every component is implementable.

Designer Skill · Quality Gate

design-qa

The final gate check before engineering gets the file — policy compliance, Vuetify compatibility, state coverage, and microcopy quality in one pass.

Designer Skill Final Gate Only Pre-Handoff
When to use

"QA this design", "is this ready for handoff", "run a final check", "check the design before I send it to engineering". Use ONLY when the design is complete — not during active design work.

What it outputs

Combined review: policy compliance + Vuetify coverage + state coverage (all user and system states covered?) + microcopy quality + handoff readiness checklist. All blockers must be resolved before handoff.

Invoke: /design-qa [feature or Figma link] — all 🔴 Blockers must be resolved before proceeding to /design-handoff

Designer Skill · Engineering Handoff

design-handoff

Generate a complete, implementation-ready handoff document that engineering can act on without needing to ask the designer follow-up questions.

Designer Skill Engineering Handoff Step 2 in /design-handoff
When to use

"Prepare the handoff", "write the handoff doc", "I'm ready to hand off this feature", "generate the spec for engineering". Requires vuetify-constraint-check to have been run first.

What it outputs

Component-level spec: each element with Vuetify component + props + variants + responsive behavior. State matrix showing all UI states. Microcopy table. Token references. Implementation notes and open questions.

Invoke: /design-handoff [feature name or Figma link] — auto-runs vuetify-constraint-check first to ensure full Vuetify coverage before generating the document

Designer Skill · Code Generation

figma-to-code

Translate a Figma design directly into Vue 3 + Vuetify 3 component code — read from Figma via MCP, not from screenshots or descriptions.

Designer Skill Reads Figma via MCP Standalone
When to use

"Generate code for this design", "turn this into Vue code", "write the Vuetify component for this", "translate my Figma to code". Typically after vuetify-constraint-check confirms all components are implementable.

What it outputs

Vue 3 SFC code with correct Vuetify 3 components, props, and responsive behavior. Follows the project's existing component patterns. Marked as reference code — engineering reviews and adapts before production use.

Reads Figma directly via MCP — no screenshots needed. Invoke: /figma-to-code [Figma file link or frame name]

Designer Skill · Content

microcopy-writer

Write clear, consistent, and on-brand UI copy for any product element — following the microcopy standards in design-policy.md.

Designer Skill Standalone
When to use

"Write copy for this", "help me write the error message", "what should this button say", "review this microcopy", "write the empty state text". Can be called during design, before handoff, or during QA.

What it outputs

UI copy for the requested element type: button labels, error messages, empty states, tooltips, confirmation dialogs, placeholder text. Includes alternatives and rationale for each choice.

Invoke: /microcopy-writer [element type + context] — covers buttons, errors, empty states, tooltips, and any in-product text

Designer Skill · Post-Implementation

implementation-review

Compare the Figma design against the actual Vue/Vuetify implementation — identify every gap and produce a structured report the designer can act on or escalate.

Designer Skill Reads Figma + Codebase Standalone · Post-Dev
When to use

"Does the implementation match the design", "check what was built against Figma", "review the implemented component", "does the code match my design". Run after engineering marks a feature complete, before the designer signs off.

What it outputs

A gap report comparing Figma spec vs. code: visual gaps (spacing, color, typography), component substitutions, missing states, interaction gaps. Each gap classified as Blocker / Minor / Accepted deviation.

Requires Figma MCP access AND repository access. Does not accept screenshots as substitutes. Invoke: /implementation-review [Figma link + component path]

Shared Skill · UI Planning

design-system-check

Inventory existing UI components before planning new design work — prevent duplicate components and make sure new requests are truly net-new.

Shared — PM & Designer
When to use

"What components do we have", "do we need to build this UI from scratch", "the designer said we need a new component", "is there already a pattern for this", before starting new UI design work.

What it outputs

Component inventory: available (can reuse as-is), extensible (needs modification), and net-new (must be built). Flags cases where a new component request could be satisfied by extending an existing one.

Invoke: /design-system-check [describe the UI needs]

Shared Skill · Visualization

wireframe-generator

Generate interactive HTML wireframes showing a feature in multiple states and multiple UX pattern variants — a concrete design starting point before high-fidelity work.

Shared — PM & Designer HTML Output
When to use

"Show me what this looks like", "generate a wireframe", "I want to see all the states", "what are the different UX options", or as Step 4 in the PM /new-feature chain.

What it outputs

A self-contained HTML file with clickable state transitions: empty state, loading, success, error, and all edge case states. Multiple UX pattern variants presented side-by-side using existing design system components.

Invoke: /wireframe-generator [feature + states to show]

Shared Skill · Linear Integration

task-writer

Convert design decisions, reviews, or fixes into ready-to-paste Linear tasks with the team's standard four-part structure.

Shared — PM & Designer Linear-Ready Output
When to use

"Write a Linear task for this design fix", "log this as a task", "create a task from this design-qa finding". Use after any review that produces actionable items that need to be tracked.

What it outputs

A four-part Linear task: (1) Context, (2) User Story, (3) DOD with clear completion criteria, (4) Out of scope. Ready to paste directly into Linear.

Invoke: /task-writer [describe the task]

Shared Skill · Meetings

meeting-support

Prepare structured agendas before design critiques and team meetings, and extract decisions and action items after them.

Shared — PM & Designer Pre & Post Meeting
When to use

"Prepare the agenda for the design critique", "what are the action items from today's review", "summarize the design decisions we made", "log the feedback from the stakeholder review".

What it outputs

Pre-meeting: timed agenda, discussion goals, and materials checklist. Post-meeting: structured summary with decisions, action items with owners, and open questions requiring follow-up.

Invoke: /meeting-support [meeting type + context]

Shared Skill · Communication

presentation-style

Generate polished, single-file HTML slide decks following the team's design system — flat-color, no-shadow, scroll-based, publication-ready.

Shared — PM & Designer HTML Output
When to use

"Make a presentation", "create slides for this", "I need to present the design to stakeholders". Works for design reviews, system proposals, process documentation, and team updates.

What it outputs

A self-contained HTML file saved to docs/ — title slide, content slides using cards/step rows/arch layers, and a summary slide. Follows the team design system: warm palette, 0.5px borders, font-weight 400–500 only.

Invoke: /presentation-style [topic and content description] — saved to docs/ for GitHub Pages

Shared Skill · Prompt Engineering

prompt-optimizer

Analyze, critique, and rewrite prompts to maximize output quality — for both claude.ai conversations and Claude Code skill invocations.

Shared — PM & Designer Prompt Engineering
When to use

"Optimize this prompt", "how can I get better results", "review my prompt", paste a prompt and ask for feedback. Works for design descriptions sent to Claude, review requests, and any claude.ai conversation prompts.

What it outputs

A critique (what's vague, missing, or over-constrained), a rewritten prompt with specific improvements, and an explanation of each change and why it improves output quality.

Invoke: /prompt-optimizer [paste your prompt]

Chain Command · Design Review

/design-review

A complete design review in one command — policy compliance check followed by Vuetify constraint check, producing a unified review report with all blockers and required actions.

Chain Command 2 Skills in Sequence Mid-Process
1

design-policy-review

Check against design-policy.md — Vuetify coverage (surface level), RTL layout, state coverage, microcopy quality, accessibility baseline. Outputs intermediate status before proceeding.

2

vuetify-constraint-check

Deep Vuetify 3 component coverage — map each UI element to a component, check theming compliance, flag gaps with implementation options.

3

Unified Review Report

Consolidated findings organized by category (policy / Vuetify / microcopy) with severity flags, required actions (🔴 Blockers), and optional improvements (🟢 Suggestions).

Invoke: /design-review [feature name + states list] — use mid-process while design is still in progress. For completed designs, use /design-qa instead.

Layer 1 · Policy

5 hard rules — always active

Designer rules enforce design boundaries: Vuetify compatibility, policy-first decisions, wireframes before high-fidelity, and authority respect.

design-policy-first Designer only
Every design recommendation is checked against design-policy.md before being presented. Policy violations are flagged, never silently ignored.
vuetify-compatibility Designer only
All UI suggestions must use Vuetify 3 components where they exist. Custom components are only proposed when Vuetify has no coverage. Always includes component name and key props.
wireframe-before-design Designer only
Claude never generates high-fidelity UI without a wireframe step first. The wireframe must be approved before design work proceeds.
flag-authority-limits Shared
Any decision outside the designer's authority is marked ⚠️ Authority with a clear statement of who must sign off. Applies to design system changes, scope additions, and new patterns.
working-language Shared
All outputs in the designer's chosen language. Component names, prop names, and Vuetify terms always stay in English. Wireframe labels use the working language.
Process · Onboarding

Designer Onboarding — Figma does the work

Two commands, two questions, and your Figma links. The design system context is extracted automatically — no copy-pasting component lists, no manual descriptions.

Prerequisites
Claude Code CLI installed Figma MCP connected in Claude Code settings Node.js 18+ & npm Figma design system file URL + product screen URLs
1

Install the package globally

One command. Installs the claude-pm CLI on your machine.

npm install -g claude-pm
2

Run init inside your product repo

Installs all designer skills to ~/.claude/ globally, and scaffolds the policy tree, placeholder context files, two design agents, and an outputs/ folder for deliverables.

cd your-product-repo
claude-pm init --role designer
3

Open Claude Code at repo root

Launch Claude Code inside the product repo. The interview command checks Figma MCP access automatically and provides setup instructions if Figma is not yet connected.

4

Run the designer interview

Type /start-designer-interview. Claude verifies Figma MCP access, then asks two questions: your working language (Persian or English) and whether to connect an existing git repository.

/start-designer-interview
5

Provide your Figma file links

When prompted, paste your Figma design system file URL and one or more product screen URLs. Claude reads the files directly via MCP — components, tokens, variants, and flows are extracted automatically.

6

Review and approve the 11 generated files

Claude shows each file before writing. Review CLAUDE.md (your designer profile), the 7 skill context.md files, and both design agents. Approve each write before Claude saves it.

7

Commit to the repo

git add CLAUDE.md .claude/ .gitignore && git commit -m "feat: add designer workflow context"
8

Run your first command

The design system context is loaded on every session. Start with a review or use /figma-to-code on an existing screen to verify the context is accurate.

/design-review [feature name or Figma link]

Refreshing context: When your Figma design system changes significantly, re-run /start-designer-interview to refresh context files. The interview only updates what changed.

Updating policy: Run claude-pm update to pull the latest skills and rules. Your Figma-generated context files are never overwritten.

Process

Receiving updates — one command

Policy Layer updates from the supervisor never touch your CLAUDE.md or context files. No re-interview, no merge conflicts.

When supervisor updates a skill or rule

Supervisor pushes to product-team-claude-skills repo. Designer runs:

cd claude-workflow && git pull && cd ..

That is the entire update process. Nothing else required.

What changes

claude-workflow/policy/skills/product-designer/*/SKILL.md
updated
claude-workflow/policy/rules/*.md
updated
claude-workflow/policy/commands/product-designer/*.md
updated

What never changes

CLAUDE.md
untouched
.claude/skills/product-designer/*/context.md
untouched
.claude/agents/design-agent.md
untouched
CLAUDE.local.md
private · gitignored

Context files extracted from Figma are designer-owned — the supervisor never touches them. Policy files and context files live in separate directory trees, so git pull never creates a merge conflict.

Summary

What this system delivers

Figma-native

Context from the source

Context isn't typed in chat — it's extracted from your Figma file. Component inventory, tokens, and variants are always accurate.

Vuetify-aware

Every suggestion is implementable

All design suggestions use Vuetify 3 components with correct props. No suggestions that engineering can't implement with your stack.

Policy-enforced

Rules run automatically

Policy compliance, state coverage, and authority limits are checked on every response. No manual checklist needed.

11
Skills available
6
Slash commands
5
Always-active rules
11
Generated context files