v2.0 Rebuilt on Claude Code dynamic workflows — a ground-up upgrade from the Cursor-agent MVP. see what's new ↓
For design leaders who run the room

Walk into every crit with the gaps already highlighted.

Superdesigner is the design review that runs before the review. It connects your PRD, research, Figma, and analytics — then hands you the missing states, broken flows, and unanswered questions. It reviews intent, not pixels, and every finding cites its source.

Get it on GitHub ↗ See how it works runs on Claude Code · pairs with Cursor
The job nobody put in your title

You’re expected to hold the whole product in your head.

PRD in Confluence research in Notion designs in Figma data in dashboards feedback in tickets decisions in Slack

In the crit, you connect all of it live — and you’re the one who’s supposed to catch what everyone missed. Superdesigner does that prep, so you never get blindsided.

01 / why leaders run it
NEVER MISS THE MARK

Every gap, mapped before you walk in

Missing states, edge cases, and unmet PRD requirements — surfaced and prioritised before the review, not discovered in it.

5 states · every screen
SCORE THE BROWNIE POINTS

Ask the sharp question first

Walk in with the redlines already done and the right questions queued. Be the person in the room who saw it coming.

questions, not verdicts
MANAGE QUALITY AT SCALE

The same rigour on every project

One consistent, senior-level review across every flow and every designer on your team — without you reviewing every pixel yourself.

pre-handoff · pre-launch · gap-audit
EVIDENCE, NOT OPINION

End the taste wars in crit

Every finding cites a PRD line, a research quote, a DESIGN.md rule, or a data point. If it can’t cite a source, it doesn’t ship.

cite the source or cut it
02 / the panel

Five reviewers in the room — every single time

Each one has a single job, its own sources, and only the tools it needs. They run in parallel, then their findings are merged, ranked, and capped to the ten that matter.

01

PRD

Does every requirement have a screen and the states to make it work?

02

Figma

Reads the file, maps the screen structure and node IDs. The only one that touches Figma.

03

UX

Missing states, cognitive load, flow clarity, trust signals — graded against your DESIGN.md.

04

Content

Copy against the brief and your voice: vague labels, legalese, missing error microcopy.

05

Analytics

Is what you want to measure actually measurable? Flags the events that won’t fire.

03 / the flow

From context to comments, end to end

Made with dynamic workflows in Claude Code — a script orchestrates the run, so the parallel fan-out, the dependencies, and the synthesis are guaranteed, not left to chance.

context

You drop in the context

Figma · PRD · research · analytics · copy brief · DESIGN.md
intent parser

What kind of review is this?

pre-handoff · pre-launch · gap-audit
orchestrator

Dynamic workflow

Plans which agents run, and in what order.

✲ made with Claude Code dynamic workflows
agent
PRD

requirements → screens

agent
Figma

screen structure

agent
UX

states & flow

agent
Content

copy & voice

agent
Analytics

measurability

tool registry
Scoped tools

Figma MCP · DESIGN.md · per-agent access

memory layer
What carries over

session · project · your preferences

synthesis

Rank, merge, prioritise

dedupe · attach evidence · rank by impact · cap to 10
guardrails

Evidence or it doesn’t ship

every finding cites a source · no taste as fact · questions, not verdicts
output
design-review.html

a self-contained, branded report

output
Figma comments

max 10 · ready to post · one idea each

observability

An auditable trail

what ran · what was skipped · what the designer overrode
04 / how it works

Three steps. About a minute of your attention.

Drop in the context

Point it at your PRD, research, Figma link, analytics, and copy brief. No design yet? It’ll propose the screens from your PRD instead.

Run the review

One command — superdesigner review — or /review inside Claude Code. Five agents fan out and reason about your intent.

Walk in prepared

You get a cited design-review.html and up to ten ready-to-post Figma comments — ranked by impact, one idea each.

05 / what lands on your desk

A review you’d be proud to have written

design-review.html — Voice Money Transfer · intent: pre-launch
P0
Is the amount shown in the exact same format at capture and at confirmation?
research §1 “if the number looks even slightly different I start over” · DESIGN.md §3 plain-language · Figma 10:220 — the documented driver of the 23% confirmation drop-off.
P0
Where is the failure state — the flow ends at Success, but the PRD requires a reason and a retry path?
PRD §4 failure state + retry · research pain #3 “it just said error” · no error screen in the Figma inventory.
P1
Does the Confirmation CTA spell out the consequence of an irreversible action?
content.md “Send AED 250 — this can’t be undone” · DESIGN.md §3.1 consequence language.
reviews intent, not pixelsevidence or it doesn’t shipmax 10 comments — signal, not noisequestions, not verdictsmade with Claude Code dynamic workflows
Before the next review on your calendar

Be the leader who already saw it.

Run it once before your next crit. Catch the gaps, ask the sharp questions, and hand your team a review grounded in evidence — not taste.