prototype-lab
Skill1 starsChanged 3 months agoMITView on GitHub
Prototype browser UIs, flip live visual variants, compare standalone options, or package a portable static build.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- lglucas/ai-dev-operating-systemtheir own version, first seen Sep 28, 2026
- dkpapadopoulos/auto-claude-skillstheir own version, first seen Sep 22, 2026
- markoblogo/abvx-agent-skillstheir own version, first seen Jun 14, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- markoblogo/abvx-agent-skills★ 16same name
- lglucas/ai-dev-operating-system★ 12same name
- dkpapadopoulos/auto-claude-skills★ 6same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 20
Read quality-bar.md, then inspect the rendered UI and exercise that path.
- Distinctive: line 27
Use the visual-axis criteria below and the bundled picker guide.
- Distinctive: line 29
Keep the requested comparison scope: a control, component or coherent screen.
- Distinctive: line 30
Default to three directions, cap five unless requested.
- Very specific: line 38
Before archive or upload, read portable-run-pack.md.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Prototype Lab
Use quick for a browser UI; picker for live visual directions of the same piece; compare for independently runnable options; ship for a static ZIP. This folder contains the full prototype workflow. One-file explanations, evidence reports and non-browser logic experiments can use the repository's normal tools; no other skill is required.
Quick: edit, observe, verify
Resolve the installed skill root and the target workspace. Call the portable tool with both paths explicit:
node "<skill-root>/scripts/manage-prototype-lab.mjs" <command> --workspace "<workspace>"- Reuse the requested or existing owner for this idea. Otherwise run
quick --title "<title>" --question "<decision>"; it creates the owner and refreshes inventory. Use its returnedidand workspace-relativepath. No separateinitis needed. - Edit inside that owner. Default
blankis an empty canvas, not a finished design. Build the composition around the task and product: real content, a clear primary action, deliberate type and spacing, and the states needed to answer the question. Keep runtime and assets local. Done when the main path visibly answers the decision. - Read quality-bar.md, then inspect the rendered UI and exercise that path. Use
preview --id "<id>" --openwhen an HTTP origin is needed; reuse a healthy preview first. Repair observed defects. Done when the relevant visual and interaction checks pass. - Run
verify --id "<id>"for known static references. It checks local HTML/CSS references and literal module imports; generated URLs and runtime requests still need browser checks. A blocked result exits nonzero. Attach an actual browser capture withattach-proof --id "<id>" --file "<absolute-image-path>"when proof is needed. Report the observed state, result and remaining limits. Static verification or a copied screenshot alone cannot prove interaction.
The neutral default blank scaffold comes from assets/prototype-blank/. Choose --profile tool|mobile|canvas only when that scaffold fits the task. adopt --path "<absolute-static-folder>" --question "<decision>" imports an existing build; fork --from "<id>" creates a distinct iteration without copying proof. Keep useful owners until the user authorizes retirement.
Picker: one piece, distinct directions
Use the visual-axis criteria below and the bundled picker guide; no external variant workflow is required.
- Keep the requested comparison scope: a control, component or coherent screen. Inspect its stack, product tokens, context and decision. Done when all directions can answer the same question with the same content and behavior.
- Default to three directions, cap five unless requested. Give each a name and a structural axis: hierarchy, layout, density, typography or interaction. Color-only changes do not count as a direction. Done when differences and their costs are visible.
- Build an isolated route or HTML file using picker.md. One direction at full size in realistic context; reserve space for picker chrome. No dated owner or lab CLI is needed unless the user also wants a catalogued artifact. Done when switching, keyboard use and reload selection work.
- Apply quality-bar.md to every direction. Show the set with one sentence per tradeoff. If selection is unresolved, let the user choose; if a winner and integration are already authorized, continue. Remove the picker only within the user's cleanup authority.
Compare and package
Compare independently runnable owners with compare --ids "<id>,<id>". Its tab and iframe switcher references the originals; each original still owns its runtime and proof.
Before archive or upload, read portable-run-pack.md. ship --id "<id>" [--include-proof] runs the static checks, writes dist/prototype-lab/<slug>/ plus ZIP, and marks the owner complete. That status means packed; it does not establish browser acceptance or publication. Public sharing retains the session's authorization boundary.
scripts/package-prototype-lab.mjs owns the portable pack. For ChatGPT Sites, use the available Sites building and hosting capabilities within the authorized publication scope.
Ownership and gallery
Each independent artifact lives at prototypes/<YYYY>/<MM>/<NNN>-<slug>/, runs from its index.html, and uses metadata.json as its source of truth. Preserve unmanaged folders.
Creation, init, sync and ship refresh prototypes/index.json. The managed generic gallery updates only while its generated content is unchanged. Existing custom or edited renderers stay byte-for-byte intact; refresh their data through the project's gallery tool. Older unmarked HTML belongs to the project. sync --renderer replace requires explicit authority to replace that renderer.
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.

