01
The win: your readiness score
Run the audit against your system in the first half hour. Get a 0–1 score and a named gap list.
Done looks like Your score, and exactly what's missing.
Workshop · 2 live sessions
Your design system, agent-ready in four hours.
Agents drift because your system was written for humans. Audit it, score it, and harness one pipeline — so an agent builds one on-brand screen through it, live, and the gates prove it.
Taught 2,000+ professionals — cohorts, fellowships, teams
Start here
What's an AI-ready design system — and what changes when yours is?
Never heard the term?
Your design system is the source your product's UI is built from: tokens, components, and the rules between them. Almost all of them are written for humans. AI-ready means an agent can read and follow yours — one source of truth, contracts it can act on, and gates that catch what slips.
GIF placeholder · an agent builds a screen through the harness, gates prove it
Already own a system?
Then you've seen the tax. An agent ships a screen, it's off-brand in a way your guidelines technically allow, and you fix it by hand. Harnessing flips it: agents build through the system instead of around it, and the gates prove the screen before you ever see it.
Agents drift wherever two truths disagree. The harness leaves one.
Either way you leave the same: an agent builds one on-brand screen through your system, live, and the gates prove it — no ad-hoc prompting allowed.
You leave with
Three artifacts, all yours, all built on your system.
- A scored agent-readiness audit of your own design system
- One working harnessed pipeline — tokens, a contract, a firing gate
- An agent-built screen that passed the gates, and the prioritized fix list for the rest
Who it's for
Bring your own design system — even a messy one. Especially a messy one.
A good fit
- Designers who own or maintain a design system, and design-systems teams
- Pairs from one org working the same system — the best version of this workshop
- Anyone with one screen an agent recently got wrong. The mess is the material
Not for you if
- There's no design system or tokens anywhere yet — encode a brand first in Brand Asset Engineering
- You want the audit done for you rather than a method you run yourself
- Your system lives only in Figma — message first; the audit works anywhere, but the live build test is strongest with a code-backed system
Bring
- Access to your design system — repo or Figma
- One screen an agent recently got wrong
- Claude Code, set up. The free Claude Code 101 pre-work covers setup if you're new to it
Agenda
Part one: audit. Part two: harness and prove it.
Session one · 2 hours
Show & tell: the audit
Watch every block run live on a real system, end to end.
Between sessions
Your turn
Run it on your own system — homework, with help in the community.
Session two · 2 hours
Bring your build: harness & prove it
Session two starts from the score and the questions you bring.
Every session is recorded
Session one · Watch it run: the audit · 2 hours
02
One source of truth
Consolidate one surface's tokens into a single source file. Delete a duplicate truth — agents drift wherever two truths disagree.
Done looks like Tokens an agent can trust.
03
Contracts and protected zones
Write the contract for one core component — props, states, do and don't. Then mark what agents must never regenerate: the optical nudges, the decided patterns.
Done looks like A contract, and a fence around your judgment calls.
Between sessions · Your turn
Run the audit and the source-of-truth pass on your own system, note where it fights you, and bring the score and the questions. Stuck points get answered in the community along the way.
Session two · Bring your build: harness and prove it · 2 hours
04
Proof gates
Wire a design lint with three rules from your own tokens. A gate that fires beats a guideline nobody reads.
Done looks like A gate that catches a real violation.
05
The build test
Have an agent build one on-brand screen through the harness — no ad-hoc prompting allowed. QA it by the gates, fix by small passes, re-run.
Done looks like A screen that passes, built from the harness alone.
06
The next three fixes
Rescore. Write the prioritized fix list so the harness keeps growing after the session.
Done looks like A plan you can hand your team on Monday.
Proof
Every method has a receipt.
- tinydesignshop.com — a storefront shipped with agents through a harness
- madhurimaram.com/harness — the harness method, published
- github.com/iruhdam1/skills — the skills, public. The audit skill you run in block 01 lives here
- What participants say — the shared wall of love, filled from cohort 1 onward
Testimonial · name · role
Testimonial · name · role
Testimonial · name · role
Instructors
Taught by people who ship this way daily.
Photo
Madhuri Maram
Lead instructor. Product designer, 2,000+ professionals trained across cohorts and fellowships, designing agentic products in production today.
Photo
Arjun
Co-instructor. Building the course material with Madhuri. Bio and receipts land on the instructors page.
Price & rooms
One price. Pick your room, then hold your seat.
$499 per seat · October cohort · live online, recorded. First room to reach eight runs; dates set at least two weeks ahead. Your pick pre-fills the WhatsApp message.
- Two live build sessions of two hours, on your own design system
- Homework and help over the community between sessions
- Recordings of every session
- The audit skill and method, public and yours to keep — plus Claude Code 101 (free) pre-work if you're new to it
FAQ
Questions, answered.
What should I bring?
Access to your design system — repo or Figma — and one screen an agent recently got wrong. The mess is the material.
Do I need Claude Code?
Yes. The free Claude Code 101 pre-work covers setup if you're new to it.
Our system lives only in Figma. Does this still work?
The audit and contracts work anywhere; the live build test is strongest with a code-backed system. Message first and we'll check fit honestly.
Can my team attend together?
Yes — pairs from one org working the same system are the best version of this workshop. Message about team seats.
Why $499?
It's the introductory-cohort price. The first cohorts shape the format, and the price says so. It rises for later cohorts.
What if the live time doesn't fit my timezone?
Every session is recorded, and the work happens between sessions on your own time. Follow the recording and bring questions to the community.
Seats
October rooms are forming.
Pick your room above — the WhatsApp message pre-fills.