damusix/atomic-claude/atomic-visual-options
Claude Code and agent skill by damusix/atomic-claude
- community
Checking that the SKILL.md file is still in the repository.
Planning-phase visual comparison aid. Renders 2-4 side-by-side variants per decision dimension as a single throwaway, self-contained HTML file and captures the user's pick as typed terminal codes (e.g. "A2 B3"). Auto-fires on phrases like "show me a few options", "mock up some variants", "let me see this side by side", "visual options for", "compare these layouts", "show layout options", "wireframe some choices". Also invoked by /atomic-plan just-in-time when a design question is genuinely visual. Scope is strictly visual choices — mockup, layout, color, spacing, hierarchy, diagram — not conceptual or text decisions, which stay in the terminal.
Install
npx vskill@latest install damusix/atomic-claude/atomic-visual-optionsSKILL.md
Render the visual in the browser; make the decision in the terminal.
<trigger>
Auto-fire on:
- "show me a few options", "show me some options"
- "mock up some variants", "mock up some options"
- "let me see this side by side"
- "visual options for …", "compare these layouts"
- "wireframe some choices", "show layout options"
- "what would X look like", "can you visualize …"
Also fires when /atomic-plan encounters a design question that passes the see-it-over-read-it gate (see below).
</trigger>
The gate (just-in-time, see-it-over-read-it)
Only render visuals when the question is genuinely visual. The test: would the user understand it better seeing it than reading it?
A question that is merely about a UI topic is not automatically visual:
- "What should the settings wizard do?" — conceptual, answer in the terminal.
- "Which of these three settings wizard layouts works?" — visual, render in the browser.
Apply this test per question, not per session. A planning conversation might have five conceptual questions and one visual one; only the visual question gets a rendered file. When the answer is "no," stay in the terminal with a plain list or prose.
Workflow
<workflow>
1. Compose
Identify the decision dimensions. Each dimension becomes a panel (e.g. Layout, Color scheme, Navigation placement). For each panel:
- Assign a letter starting at A (panel A, panel B, panel C…).
- Generate 2–4 options. Label each with the panel letter + option number:
A1,A2,A3. No more than four options per panel — beyond four the comparison stops being useful. - Assign each option a short, descriptive title (e.g. "A2 — Two-column with sidebar").
2. Write
Write ONE self-contained HTML file to the scratchpad bundle:
Paths come from atomic scratchpad / atomic where --json; if what you find on disk does not match, run atomic migrate --show-log for the change history.
The file must satisfy these constraints:
- Starts with
<!DOCTYPE html>. - All CSS inline in a
<style>block — no external stylesheets. - No external requests: no CDN links, no web fonts loaded from a remote host, no remote images. Default to CSS geometry and SVG for wireframe placeholders so the file renders fully offline. If a real raster image is genuinely required for fidelity, embed it as a base64
data:URI — the file must remain self-contained and work without a network connection. - No client-side JavaScript. The selection is typed in the terminal; the page is pure display.
- Honors
prefers-color-schemefor light and dark mode via a@media (prefers-color-scheme: dark)block.
3. Hand off
Print the file:// path to the file. Do not auto-open the browser. End the turn.
Tell the user: open the file, pick one code per panel (e.g. A2 B3), and reply with those codes.
4. Read the pick
Next turn, parse the codes the user typed. Accept space- or comma-separated, one code per panel. Map each code back to its option title and description.
5. Iterate
If the user wants a panel changed or wants to see a different variant, overwrite the same file and tell them to refresh the browser. No versioning — the file is throwaway scratch.
6. Record
Inside a `/atomic-plan` session: write the chosen codes and what each option meant into docs/design/<topic>.md. Example entry:
Standalone (not inside a plan): state the chosen codes and what each option meant directly in the terminal so the decision is captured in the conversation history. If a docs/design/<topic>.md for the topic already exists, append the decision there too.
The HTML file is disposable and will be cleaned up with the scratchpad. The decision is what persists.
</workflow>
HTML scaffold reference
Adapt the following structure for each new options file. It is intentionally compact while remaining genuinely functional — two panels, two or three options each, a usage instruction header, inline CSS with light/dark support, and a large readable code badge per option. Replace the SVG placeholders and descriptive text with content appropriate to the specific decision.
<constraints>
Rules
- Visual questions only. Route conceptual and text choices — anything the user would understand as well by reading a list — to the terminal instead. A
<question>that happens to mention a UI component is not automatically visual. Why: the browser is a context switch; it earns its cost only when the content genuinely cannot be conveyed in prose.
- The file is throwaway scratch; never commit it. Record the chosen codes and what each option meant in
docs/design/<topic>.mdinstead. Why: a committed mockup rots and misleads future readers; the chosen code line (A2 B3) plus a one-sentence description is the durable record.
- Self-contained and offline: no external CSS, no CDN fonts, no remote images by default, no client-side JavaScript. Why: the file must render anywhere a browser exists with zero dependencies and zero setup; the selection mechanism is the terminal, so client-side interactivity is not needed.
- 2–4 options per panel, maximum. Why: beyond four, a side-by-side comparison stops being useful — the user is no longer comparing, they are scanning a catalogue.
</constraints>
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingcontext/skills/atomic-visual-options/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/damusix/atomic-claude/atomic-visual-options)