Specialization kata · UX-R01

UX Randori: an unseen flow with a hidden constraint

An unseen-constraint UX Randori: map, wire and check a small flow you have not seen, and adapt when a partner reveals one constraint after you start.

Draft. This recipe is a draft. Its checks may not have been run, and it can change.

Status
Draft
Kata ID
UX-R01
Maintainer
UX pathway maintainer
Last verified
Not available. Built and checked by the Dojo maintainers; no independent review yet.
Found a problem?
Report it

The outcome

In one session, map and wire a flow you have not seen, then adapt it to one constraint a partner reveals after you start.

For: Learners who have done UX-001 to UX-004.

Done when

  • The sitemap has no unreachable state and no dead end
  • The prototype reaches every transition
  • The hidden constraint is met, and you can say what it changed
  • The partner walk is written up: where they went, where they stopped

You make: Sitemap, Wireframe set, Qa report. Outcome: Map user flow, Connect wireframes, Implement and qa screens.

Showcase and provenance

The Dojo’s worked example, made by the maintainers rather than by a learner: Pocket Invoice as an editable Penpot file, a token file and a working, accessible page.

See the showcase

Made by
Developer Dojo maintainers
Source
Generated from the Pocket Invoice flow and tokens with Penpot’s official library
Licence
MIT
Date

Brief, inputs and constraints

Inputs

  • An unseen brief and a sealed constraint. Source: Chosen by your partner from the list in the download. Terms: Dojo content, MIT licence.
  • Moderated usability testing. Source: gov-uk/moderated-usability-testing. Terms: Open Government Licence v3.0 unless it says otherwise.

Constraints

  • Fictional data
  • No personal data
  • No paid tools

Capabilities practised

  • Map flows
  • Design interaction
  • Test accessibility

Stack, permissions and cost

Stack components with version, permissions and cost
PartComponentVersionPermissionsCost
ToolPenpot (self-hosted or penpot.app)2.18 (the example file was checked on a self-hosted 2.18.7)Edit one practice project of your ownPenpot Professional plan: $0, up to 8 members and 10 GB, no file limits (penpot.app/pricing, checked 2026-11-20). Self-hosting is free under MPL-2.0; you pay for your own server.
HarnessAny chat assistant in a browserNot available. Not fixed; any assistant can draft the text, and its version changes the wording, not the checksReads only what you paste into it; it never touches the design fileNot available. Depends on your assistant and plan; not measured
InstructionsThe prompts on this pageDraft, 2026-11-20NoneNone
DataPocket Invoice, an invented invoicing appFlow spec of 2026-11-20: 11 states, 19 transitionsFictional; no real company, person or customerNone

Stack recipes

  • Penpot and browser
  • Assistant and browser

Why this stack

Penpot is open source, runs in a browser, can be self-hosted and has a free plan with no file limits, so every learner can open the same editable file. The assistant only drafts text you check; the design work and the checks stay with you.

Alternatives

  • Figma. Widely used. The free Starter plan allows unlimited drafts but is limited elsewhere (for example 20 MCP tool calls a month); Professional seats cost $3 to $16 a month and there is no self-hosting (figma.com/pricing, checked 2026-11-20). The example file is Penpot-only and has not been tried in Figma. Not tested for this recipe.
  • Penpot MCP server with an agent. Lets an agent edit the open page directly, but uses a personal key with no permission scopes. Not run for this pathway: no native write by an agent was verified. Not tested for this recipe.

Agents and review gates

Not available. A Randori is done without an assistant, so the learner shows their own judgement.

Setup and smoke test

  1. Your partner picks one brief and one constraint from the download and keeps the constraint hidden.
  2. Open an empty Penpot project.

Sign-in and access: Your own Penpot account.

Smoke test

Do
Your partner reads the brief aloud.
Expect
You can name the user, the job and the first screen.
Result
Not run Nobody has recorded running this check. It is not a pass.

Prompts

Not available. This recipe has no prompts yet.

Practice exercise

Moderated usability testing (GOV.UK Service Manual). From Using moderated usability testing (GOV.UK Service Manual).

Terms: Read on the publisher’s site. GOV.UK guidance is published under the Open Government Licence v3.0 unless it says otherwise.

Execution and handoffs

  1. Partner: Reads the brief aloud and keeps the constraint hidden.
  2. Learner: Maps the sitemap, with errors and the way back. Hand-off: The partner reveals the constraint.
  3. Learner: Changes the sitemap and draws low-fidelity screens that meet the constraint, then wires them.
  4. Partner: Walks the prototype without help, as in a moderated test, and notes where they stopped.

Tests, failures and corrections

Checks with expected and recorded results
CheckExpectedResultEvidence
Partner walk of the prototypeNo dead end; the constraint is metNot runNot run, so there is no evidence. This is not a pass.

Failures and corrections

No failures have been recorded, because no check has been run.

Adapt it

  • Two constraints instead of one. Harder: they may conflict, and you must say which one wins and why.

Unseen Randori

Map and wire an unseen flow, then adapt it to a constraint your partner reveals only after your first sitemap.

Constraints

  • The constraint is unseen until the first sitemap is drawn
  • Fictional data only
  • Every error state has a way back
  • No assistant during the session

How it is judged: The partner walks the prototype without help. Pass: no dead end, the constraint is met, and the learner can say what the constraint changed.

More Randori challenges

Belt evidence

Belt stages are proposed practice stages, not issued credentials. Evidence below is what a reviewer would look for; it is not a record that anyone has shown it.

Evidence a reviewer looks for

  • A partner who did not see the work walks the sitemap and the prototype and records where they stopped (Peer review)
  • The learner records what the hidden constraint changed, with the before and after (Self-check)

Belt criteria

  • Demonstrate: Delivers a sitemap and a wired low-fidelity prototype for the unseen flow that meet the hidden constraint, with a partner walk that finds no dead end

See the belt stages · Prepare a practice review

Connections

Used by
No other kata names this one as a foundation yet.
Professional review
Not required for practice. Real work in a regulated field still follows that field’s rules.

Sources

Downloads

Plain text files. Read them before you use them; nothing runs when you download one.

  • randori-cards.md: Briefs and hidden constraints for the partner. The learner should not read it first.

Maintenance and caveats

Review cadence: Every six months, or when Penpot changes its file format or token import

  • This recipe is a draft: the example was built and checked by the Dojo maintainers, and no learner has run it yet.
  • Penpot plans, versions and features change; check penpot.app/pricing and help.penpot.app before you rely on a limit.
  • The example file was checked on a self-hosted Penpot 2.18.7. Penpot cloud (penpot.app) was not checked.

All kata recipes · Back to katas