agentleFS
Evan-Kim2028Evan-Kim2028/agent-skills/skills/frontend/ui-explore/SKILL.md

ui-explore

Skill1 starsChanged 3 months agoMITView on GitHub

Throwaway design exploration before production — standalone HTML A/B pickers, in-app multi-variant UI prototypes, or tiny terminal logic prototypes. Use when the user wants design options, compare A vs B, "prototype this", "try a few designs", or /html-design. Prefer product-design when multi-step product UX craft is unclear; frontend-design for full FE builds. Not for production ports (mockup-implement after pick), charts (tufte), marketing copy, or ship proof (quality-check / browser-verify). Aliases: html-design, prototype.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    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
  1. Distinctive:

    Throwaway artifacts that answer a design or logic question: before committing production.

    line 15
  2. Distinctive:

    Prefer HTML A/B for small product deltas and review without a full app boot

    line 34
  3. Very specific:

    Prefer in-app variants when judgment needs real chrome/data density → references/ui-branch.md + HTML rules below

    line 35
  4. Distinctive:

    If ambiguous and user is unreachable: product page → UI.

    line 38
  5. Distinctive:

    Surface state: after every action / variant switch.

    line 46
  6. Distinctive:

    Delete or absorb: when the question is answered.

    line 47
  7. Distinctive:

    Verify switcher; user picks.

    line 68
  8. Very specific:

    Prefer embedding variants on an existing route with ?variant= (real header/data).

    line 77

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

UI explore

Throwaway artifacts that answer a design or logic question before committing production. One skill, three shapes.

Sources (attribution)

  • HTML A/B: Anthropic The unreasonable effectiveness of HTML; local design-picker patterns (?design=, floating switchers).
  • In-app / logic prototypes: throwaway prototype branch pattern (UI variants + terminal state machines) — original synthesis; templates in-repo.
  • Full pack credits: ATTRIBUTION.md.
ShapeQuestionDetail
HTML A/B“Which pixels for this product delta?”Standalone HTML + picker (default for product chrome deltas)
In-app UI variants“What should this look like in the app?”?variant= on a real route + floating bar
Logic / terminal“Does this state model feel right?”Tiny TUI over a pure reducer/machine

Pick a shape

  1. Logic / state / API feel → references/logic-branch.md
  2. Look & layout
    • Prefer HTML A/B for small product deltas and review without a full app boot
    • Prefer in-app variants when judgment needs real chrome/data density
      → references/ui-branch.md + HTML rules below

If ambiguous and user is unreachable: product page → UI; backend module → logic. State the assumption at the top of the artifact.

Rules for every shape

  1. Throwaway and marked as such (banner, path name ui-explore, title).
  2. One command to run (project runner or python3 -m http.server).
  3. No production polish (no full test suite, no gold-plating).
  4. Surface state after every action / variant switch.
  5. Delete or absorb when the question is answered; capture the verdict.
  6. Live data when the surface already has a backend. Bind to the real API or a recorded snapshot of it so density is honest. Empty fixtures hide the product. Do not rewrite production chrome to explore — product-design principle 9. If web / mobile-web / native share the surface, the variant must be judgable on all of them (or say which surface is out of scope).

Shape 1 — Standalone HTML A/B (product deltas)

Inspired by The unreasonable effectiveness of HTML.

Hard rules

  1. Match blast radius. Small ask → Design A = current, Design B = current + only that delta.
  2. HTML first, production second.
  3. One switcher surface (floating picker / keys / ?design=).
  4. Serve + verify the URL; give the user the link.
  5. Record the pick (one line: which design, what to implement).

Workflow

  1. Frame one sentence question.
  2. Mode: A/B before–after (default) or multi-variant (max 5).
  3. Build one HTML file — scaffold templates/ab-compare.html; patterns references/patterns.md; anti-patterns references/anti-patterns.md.
  4. Serve (./design/serve.sh or python3 -m http.server 8765).
  5. Verify switcher; user picks.
  6. Implement via mockup-implement only after sign-off.

Paths: design/<topic>-ab.html when the repo has design/.

Shape 2 — In-app UI variants

Full process: references/ui-branch.md.

  • Prefer embedding variants on an existing route with ?variant= (real header/data).
  • Floating bottom bar; 3 variants default, max 5.
  • Mark route/file as prototype; delete or fold after decision.

Shape 3 — Logic terminal prototype

Full process: references/logic-branch.md.

  • Pure logic module + throwaway TUI.
  • One command; print full state after each action.
  • Lift validated logic into production; delete the TUI.

Done criteria

  • User can run one command / open one URL
  • They can switch variants or drive cases without re-prompting
  • Verdict captured (which option / what we learned)
  • Production path named or explicit reject + delete plan

Hand off

NextSkill
Port signed-off visualmockup-implement
Polish after portproduct-ui-craft (+ mobile-product-ux)
Tokensdesign-system
Prove in browserbrowser-verify / quality-check

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.