first-mocks

Turns a concept into the cheapest concrete artefact a human can react to, via the no-code make-sequence: Crazy 8s → paper/Miro flow → digital mock → clickable hybrid in v0 or Figma Make, built in one

SKILL.md

GitHub

What it does

Takes a concept you can only describe in words and forces it down to the cheapest artefact a human can actually react to. It runs the make-sequence — Crazy 8s → paper/Miro flow → digital mock → clickable hybrid — and stops at a clickable prototype built in one afternoon, no code. The output is a Clickable-Prototype Plan (template.md): the ONE moment the mock must provoke a reaction to, the fidelity band that reaches it, the hard no-code ceiling, where the visual craft gets routed, and the hand-off to a real reaction. It does not draw the pixels and it does not run the test. Its job is to convert an abstraction into something concrete, fast, and disposable — then get out of the way.

The Icarus reframe

A mock is not a small version of the product. It is the cheapest concrete thing a human can react to, because people give useless feedback on a description and useful feedback on an artefact. So fidelity is set by one question — what is the least-built thing concrete enough to provoke a real reaction to the ONE moment I am unsure of — under a hard ceiling: no code, one afternoon. Code is banned here because it converts a disposable question into a defended commitment, and because the Crazy 8s exist to force divergence that a code editor kills the moment you start building the first idea you had. The generic fidelity ladder keys the choice to how much you want to build; this keys it to how little you can build and still get a real reaction.

When to use / When NOT

Use it when you have a concept and need a concrete artefact people can react to, before you write any product code. It sits after you know what you are testing and before anyone watches a user or ships a line.

Do not use it for these — hand off instead:

If the fellow wants… Use instead Why not this skill
Which cheap experiment answers their ONE learning question (paper / WoZ / concierge / agent) probe-matrix That selects the experiment and its honesty; this MAKES an artefact. Route there first if the question is "which test."
The paper sketch run as an evidence-gathering probe paper-sketch-probe That runs a sketch as a probe with a read-out; the paper step here is one rung in a make-sequence, not the probe.
To watch a user react to the mock and score it usability-test-protocol That runs the session and captures behaviour (0.7); this produces the thing the session uses.
The visual / UI craft — layout, colour, type, distinctive non-generic components impeccable / shape / design/* Those own the look and do it better. This routes to them and never restates them — restating one is the kill line.
To write the real product code compound-system-architecture, eval-first-spec (07) The moment you write product code you have left first-mocks. Code is banned at this stage.

This skill owns the sequence and the anti-code discipline. It does not choose your experiment, run your test, style your pixels, or build your product.

Method

Fill in template.md. One question, then four make-bands, then two hand-offs.

Step 0 — Name the ONE thing the mock must provoke a reaction to. One moment, one flow, one decision. Write the reaction you are hunting: "can they see the next action cold", "which of these two flows confuses less", "does this moment read as X." A mock answers legibility / comprehension / flow questions (does the artefact make sense). It does NOT answer will-they-pay or would-they-use-it — those need a probe (probe-matrix), because a mock tops out at "artefact shown" (0.5). If Step 0 is blank you will build a demo of everything and learn nothing. If Step 0 is a solution ("mock the mobile app"), it is a solution-in-disguise — reframe to the job first (problem-statement), then return.

Steps 1–4 — The make-bands. Climb only as far as the reaction needs. The ceiling is fixed at each band; breaking it is the failure named in the last column.

Band Artefact Concrete enough to provoke… Timebox Ceiling — do not cross
1 Crazy 8s 8 rough frames, one moment each your own divergence; which directions even exist 8 min pencil only; no favourite yet
2 Paper / Miro flow one flow: screens as boxes, actions as arrows is the path legible to you / a colleague ~30 min no pixels
3 Digital mock static frames (Figma / slides / Miro) can someone read the moment cold 1–2 hr no code
4 Clickable hybrid frames wired to click the ONE flow (v0 / Figma Make) can a hand click it and react one afternoon (≤ ~4 hr) no backend, no real data, no auth, no code you'd keep
— product code nothing new at this stage — BANNED here — belongs to build (07), gated by v1-launch-bar

Step 1 — Crazy 8s. 8 frames, 8 minutes, one screen or moment per frame. The first frame is a reflex; the value is in the strange 6th–8th. Keep the odd child. Fewer than 4 genuinely different directions means you decorated your first instinct — redo it cold. Do not skip to the layout you already had in mind. If real divergence already exists — you arrive with several genuinely different rough layouts — band 1 is already satisfied; converge them at Step 2 instead of forcing a fresh eight-frame round. The bar is divergence, not the ritual.

Step 2 — Converge to one flow on paper / Miro. Pick the clearest through-line, not the prettiest frame. Lay the flow: boxes for screens, arrows for the actions that move between them. This skeleton is what the clickable hybrid will wear. Still no pixels.

Step 3 — Digital mock, no code. Turn the paper flow into static digital frames. This is where the code editor calls. Do not answer it. The mock is disposable; code is not, and the moment it exists you start defending it.

Step 4 — Clickable hybrid in an afternoon. Wire the frames in v0 or Figma Make into something a hand can click through the ONE flow. "Hybrid" means real enough to click, faked everywhere else — no backend, no real data, no auth. The scaffold v0 or Figma Make emits to make the frames clickable is NOT the banned "code": the ban is on hand-writing product code you would keep and defend. A disposable generated scaffold you would delete without a pang is fine; a backend you start maintaining is not. Timebox one afternoon. If it runs into days you are building the product, not mocking it; stop.

Step 5 — Route the visual craft OUT. Do not polish pixels here. Hand the flow to impeccable / shape to plan the UX/UI, and impeccable to build distinctive, non-generic UI, or design/* for a specific surface. first-mocks owns the sequence and the no-code discipline; the look is theirs. Restating their layout / colour / type advice inside this skill is the kill line.

Step 6 — Hand to the reaction, pre-registered. The mock is an input, not an answer. Route to usability-test-protocol to watch a user react (behaviour, 0.7), or back to probe-matrix if the ONE thing needs a probe. Write the reaction that counts as a pass BEFORE anyone sees the mock. The mock produces 0.5; the reaction is where the 0.7 you actually want gets made.

Evidence standard

A mock is an artefact shown — 0.5 on the ladder (money 1.0 → behaviour 0.7 → artefact 0.5 → verbal 0.3 → opinion 0.1) — and no higher. It cannot tell you they want it (behaviour, 0.7) or will pay (money, 1.0). Its only job is to convert an abstraction into something concrete enough that a downstream test produces a real 0.7 / 1.0 read-out. Done means: a clickable hybrid, built in one afternoon, no code, visual craft routed out, and a pre-registered reaction handed to usability-test-protocol or probe-matrix. Tag every empirical claim [Fact], [Assumption], or [Hypothesis]; a mock's look and feel is at most [Hypothesis] until a user reacts to it. Kill line — the moment the mock becomes product code, or the moment this skill restates a design skill's craft instead of routing to it, it has failed.

Gotchas

  • Code is the seductive shortcut and the trap. v0 and Figma Make blur into a real app, and "I'll just wire the backend" turns a disposable mock into a defended commitment you now argue for instead of about. The ceiling — no backend, no data, no auth, one afternoon — is the whole discipline. If you would be sad to delete it, you built too much.
  • Skipping the Crazy 8s to the idea you already had. The 8 frames exist to kill your first reflex. One or two variations is decoration, not divergence; if every frame is a tweak of the first, you diverged nothing. Redo cold, and force yourself past the obvious to the strange frames where the non-generic idea lives.
  • Polishing pixels inside first-mocks. The pull to make it pretty here is strong, and it is impeccable's job, done better. Writing colour, type, or layout guidance into this output is the kill line — route it out. Keep first-mocks about the sequence and the concrete-reaction logic.
  • Reacting to the mock as if it were evidence. A polished clickable prototype feels like validation. It is 0.5, an artefact shown; no one used it under stakes and no one paid. When a fellow says "people loved the prototype," ask what they DID, not what they said — the mock is the setup for the test, not the test.
  • A mock aimed at everything. With Step 0 blank the fellow builds a demo of the whole product and learns nothing clean, because no single reaction is isolated. One moment, one unknown, one mock.

Examples

examples/sample.md — Mentix mocks the copilot's shift-handover card. The ONE thing is whether a plant manager, mid-shift, can read the card cold and act on the flagged line. The run diverges with Crazy 8s (keeping the odd "one sentence + expand" frame), converges on paper, wires four frames into a clickable hybrid in v0 in an afternoon with clearly-illustrative data, routes the look to impeccable, and hands a pre-registered comprehension read-out to usability-test-protocol. The mock is logged at 0.5; the read-out is where 0.7 is made.

Related skills

Absorbs discovery/prototyping-pretotyping — it keeps the useful bones (match fidelity to the question, throw the prototype away, behaviour over opinion) and reframes them around one axis the old skill never names: the least you can build and still get a real reaction, under a no-code afternoon ceiling. It does not supersede it; probe-matrix already does, on the experiment-selection axis. first-mocks takes the other half — the make-craft — and stays disciplined about not being the test.

Composes with: probe-matrix (selects the experiment; run it first if the question is "which test," and return here to make the artefact a paper-sketch probe needs), usability-test-protocol (runs the session that turns this 0.5 mock into a 0.7 read-out), impeccable / shape / design/* (own the visual craft this skill routes to and never restates), and the build skills in section 07 (compound-system-architecture, eval-first-spec, gated by v1-launch-bar) that own the product code this skill bans.