Structured Design Critique System
A persistent, multi-round design critique pipeline: scope which of 19 design areas your project actually needs, generate real HTML wireframe options for each, collect Keep/Cut feedback, and let locked decisions constrain everything downstream.
Every design review you've gotten from an AI is the same shape: one pass, generic feedback, no memory of what you decided five minutes ago. By the time you're picking a color palette, it's already forgotten that you chose a keyboard-first navigation model and generates decorative options anyway.
Who it's for: founders and solo developers designing a product without a design team, product designers wanting structured critique with real wireframe options instead of text feedback, teams needing a repeatable design review process across multiple projects, anyone who wants design decisions to actually build on each other instead of being independently generic, indie hackers who need something better than a single unstructured 'review my UI' prompt
Example
"Let's do a design review for my note-taking app" → A confirmed 10-area plan (screen inventory through accessibility, in dependency order), real interactive HTML wireframe options generated for each area with Keep/Cut/Comment feedback rounds, a persistent dashboard tracking progress across sessions, and later areas — like color and typography — automatically respecting the keyboard-first, minimalist navigation already locked in an earlier round
CLAUDE.md Template
New here? 3-minute setup guide → | Already set up? Copy the template below.
npx ccplaybooks add structured-design-critique-system# Structured Design Critique System
## Your Role
You lead a structured, multi-round design critique — not a single-pass "here's my feedback" review, but a persistent pipeline: brief → plan which design areas matter for this specific project → work through each area with generated option wireframes and real feedback rounds → synthesize everything into a coherent design direction.
State persists across sessions in `.design-crit/`. This can run over days, picking up exactly where it left off.
**User-facing language matters.** Always say "design area" or "area" — never say "facet" out loud, even though that's the internal term. The user should never see internal machinery vocabulary.
---
Get new playbooks like this one
One email a week with new Claude Code workflows. Free, like everything here.
No spam. Unsubscribe anytime.
What This Does
Most AI design feedback is one pass and stateless — you get a list of suggestions, and the next question starts from scratch with no memory of what you just decided. This is different: a persistent pipeline that scopes which design areas your specific project actually needs (out of 19 pre-built options, dependency-ordered), works through each one with real generated wireframe options instead of text descriptions, and carries every locked decision forward so later areas actually respect earlier ones.
State lives in .design-crit/ and survives across sessions — you can pick up a review three days later and it summarizes exactly where you left off before continuing.
Quick Start
Step 1: Create a Project Folder
mkdir design-critique && cd design-critique
Step 2: Download the Template
Click Download above, then:
mv ~/Downloads/CLAUDE.md ./
Step 3: Start a Review
claude
Then say: "Let's do a design review for [your project]" — expect a brief conversation to scope the project before any design work starts.
The Pipeline
Brief → Review Plan → Design Reviews (per area, in dependency order) → Design Direction
Design Area Library
| Core areas (default in) | Situational areas (include when relevant) |
|---|---|
| Screen Inventory, Edge States | Onboarding Flows (consumer apps) |
| Navigation, Content Layout | Data Visualization (dashboards) |
| Content Hierarchy, Component Design | Responsive (cross-device) |
| Elevation & Shape, Typography | Iconography (icon-heavy UI) |
| Color System, Density & Spacing | Motion (polish-focused products) |
| Content Voice, Accessibility (always last) | Imagery & Illustration (marketing/brand) |
| Loading & Performance (async-heavy apps) |
Typical project: 8-12 areas. A custom area gets added when the brief reveals something none of the 19 pre-built ones cover.
Tips & Best Practices
- Generate real wireframes, not descriptions. Every option is an actual interactive HTML file you can open and click through — "here's what option A might look like" as prose defeats the entire point of a visual review.
- Every option needs an honest downside. The "Watch out for" section on each option card is not optional — a review where every option sounds equally good isn't actually helping you decide.
- Later areas must respect earlier locked decisions. This is the core value proposition over a one-shot review: if navigation locked on a keyboard-first, minimal approach, the typography and color options generated afterward should reflect that constraint, not ignore it.
- The "Emerging Preferences" summary is worth reading explicitly. After 2+ areas lock, the system distills cross-cutting patterns — "consistently prefers functional over decorative" — into a running summary. That's your actual design voice being surfaced, not just a decision log.
- Skip areas deliberately, don't force every one. If a brand guide already exists, color and typography may be substantially pre-decided — the planning stage should exclude or de-prioritize areas the brief has already answered, with a stated reason.
- Re-entry summarizes before resuming. Coming back to an in-progress review should always get a status recap first — how many areas are locked, what round you're on — never a silent jump straight back into wireframe generation.
Limitations
- Designed around generating and viewing real HTML files locally — works best in an environment where you can actually open and interact with the generated wireframes, not a chat-only context
- The area-selection step requires genuine judgment tied to the brief; treating the 19-area library as a fixed checklist to run through regardless of project defeats its purpose
- Best suited for genuinely multi-decision design work (a new product or a significant redesign) — a single quick UI tweak doesn't need the full pipeline overhead
- State persistence assumes a consistent working directory across sessions — moving the project or losing the
.design-crit/folder loses the accumulated context and locked decisions