logic-prototype
Skill2 starsChanged 3 months agoMITView on GitHub
Build a single-file HTML demo that answers whether a state model, business logic, or data shape feels right — a pure liftable module driven by free-play buttons and tabbed guided walkthroughs a non-developer can click through. Use when the user wants to sanity-check a state machine or reducer, probe an edge case ("what happens if X then Y"), feel out a data model or API shape before building it, or hand a domain expert something to poke at. For what a UI could look like use prototype; for editable Paper artboards use paper-prototype.
In this repository12 other files it gives its agents.
- buildSkill
- decision-coachSkill
- grill-with-docsSkill
- handoffSkill
- harness-configSkill
- paper-prototypeSkill
- polish-fixSkill
- post-build-reviewSkill
- pre-build-reviewSkill
- prototypeSkill
- wrapupSkill
- writing-for-agentsSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
If the question is "what could this look like", wrong skill — use prototype.
- Distinctive: line 20
Write down what state model and what question this prototype answers — one paragraph, visible at the top of the demo, not a code comment.
- Very specific: line 37
Keep it pure: no DOM, no document, no button handlers reaching inside.
- Very specific: line 63
Keep it restrained: clean typography, generous spacing, one accent colour, no animation.
- Very specific: line 77
When the question is answered, add a one-line verdict (what was settled and why) to a notes file beside the HTML, and leave a pointer …
- Distinctive: line 90
Never ship the shell: Only the module lifts into the codebase.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Logic prototype
A logic prototype is throwaway code that answers one question about a state model — the kind of model that looks reasonable on paper but only feels wrong once someone pushes it through real cases. The artefact is a single, self-contained HTML file anyone can double-click: a designer, a PM, a domain expert. It speaks their language, not the code's.
If the question is "what could this look like", wrong skill — use prototype.
Process
1. State the question
Write down what state model and what question this prototype answers — one paragraph, visible at the top of the demo, not a code comment. A logic prototype that answers the wrong question is pure waste; making the question explicit lets it be checked later, whether the user is watching now or returning to it AFK.
2. Isolate the logic in a pure, liftable module
The logic under test lives in one <script> block as a small pure module that
could be lifted out and dropped into the real codebase later. The page around
it is throwaway; this module is not. Pick the shape that fits the question:
- Pure reducer —
(state, action) => state, for discrete events over a single value. - State machine — explicit states and transitions, when "which actions are even legal right now" is part of the question.
- Set of pure functions over a plain data type, when there is no implicit current state.
- Class with a clear method surface, when the logic genuinely owns ongoing internal state.
Keep it pure: no DOM, no document, no button handlers reaching inside. The
page calls into it; nothing flows the other way. That boundary is what makes
the validated module lift straight into the real code once the question is
answered.
3. Build the shareable HTML file
One file, plain HTML/CSS/JS — no framework, no bundler, no server, everything inline, so it opens by double-click and survives being sent around. Write it for a non-developer: every label in domain language, buttons and state reading like the business, not the reducer. Top to bottom:
- Title and the question from step 1.
- Current state — the full relevant state as a readable panel (labelled fields, not raw JSON), re-rendered after every click, with a callout of what just changed.
- Free-play buttons — one real button per action, always available, so anyone can poke at the model in any order.
- Guided walkthroughs — scenarios, one per tab: a short plain-language description of the situation and what to watch for, then the ordered buttons to press. Each step is a live button; starting a walkthrough resets to a known initial state so the scenario replays deterministically.
Choose scenarios that demonstrate the awkward cases: the happy path, a tricky edge, an attempt at something that should be illegal.
Keep it restrained: clean typography, generous spacing, one accent colour, no
animation. Real <button> elements, keyboard operable — nothing competes with
the state and the buttons.
4. Save, open, hand over
Save to a durable notes or artefacts location outside the repo — wherever the project keeps design records (create it lazily) — print the path, and open it. The interesting moments are "wait, that shouldn't be possible" and "I assumed X would differ" — those are bugs in the idea, which is the point. Add actions or scenarios on request; prototypes evolve.
5. Capture the answer
When the question is answered, add a one-line verdict (what was settled and
why) to a notes file beside the HTML, and leave a pointer wherever the
implementation is tracked. The validated module lifts into the real code
through the build skill; the HTML shell stays behind in the artefacts
location as the record of how the question was answered.
Anti-patterns
- No tests. A prototype that needs tests is no longer a prototype.
- No real database. In-memory state, unless persistence is the question.
- No generalising. No "what if we later want X" — one question per prototype.
- No blurring module and page. A module that touches the DOM is no longer liftable.
- No framework, bundler, or server. One double-clickable file; a dev server defeats "shareable".
- Never ship the shell. Only the module lifts into the codebase; the page stays in the artefacts location.
Credits
Derived from Matt Pocock's MIT-licensed prototype skill (LOGIC.md branch,
github.com/mattpocock/skills); upstream
licence in LICENSE-upstream.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

