unship
Compare AI agent-made UI variants locally in a real app, then keep one and clean up unused temporary code.
Best use case
unship is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Compare AI agent-made UI variants locally in a real app, then keep one and clean up unused temporary code.
Teams using unship should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/unship/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How unship Compares
| Feature / Agent | unship | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Compare AI agent-made UI variants locally in a real app, then keep one and clean up unused temporary code.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Unship
## Overview
Unship is a local workflow for comparing AI-generated UI alternatives in the real application instead of accepting one generated version at a time. It adds temporary source-level variants, shows a local browser picker, and then cleans up the unused options after the user chooses.
This skill is for frontend iteration with coding agents. It is not production A/B testing, analytics, feature flagging, or a hosted experiment service.
## When to Use This Skill
- Use when the user wants to compare multiple UI, layout, copy, state, flow, or design-system alternatives.
- Use when a coding agent should create several temporary options in real source code and let the user judge them in the running local app.
- Use when the user chooses a visible option and wants the losing temporary code removed before shipping.
## Do Not Use This Skill When
- The user needs production experiments, traffic splitting, analytics, or feature flags.
- The app cannot safely render inactive hidden variants because of duplicate active IDs, global scripts, analytics triggers, focus traps, destructive actions, or autoplay side effects.
- The user has not authorized local source edits.
## How It Works
### 1. Install or reuse Unship
Prefer the project-local binary when it exists:
```bash
./node_modules/.bin/unship doctor --json --no-update-check
```
Otherwise use the npm package without assuming a global binary:
```bash
npx -y @unship/cli@latest doctor --json --no-update-check
```
If setup is needed for the local picker, run:
```bash
npx -y @unship/cli@latest setup --json
```
Patch only the smallest development-only mount point required to load the picker in the local preview.
### 2. Create temporary variants
Inspect the relevant page, component, route, or rendered artifact. Add the smallest source-level comparison that lets the user judge real options in context.
Use Unship markup:
```html
<section data-unship-pick="Hero">
<div data-unship-option="Current">...</div>
<div data-unship-option="Proof-led" hidden>...</div>
<div data-unship-option="Visual" hidden>...</div>
</section>
```
Keep option labels short and visible. Prefer 2-4 meaningful alternatives unless the user asked for a specific count.
### 3. Verify comparison readiness
Before handing off to the user, check that:
- the expected `data-unship-pick` group exists;
- the expected option labels exist;
- options are direct children of the group;
- exactly one option is initially visible;
- hidden inactive options remain hidden.
### 4. Let the user choose
Tell the user the group label, option labels, setup status, and any detected local preview server hints. The user chooses by naming a visible option label in chat.
### 5. Clean up after selection
When the user picks a winner, keep that option's real source and remove losing options for that group. Remove temporary `data-unship-*` attributes from settled source.
For final cleanup before shipping, remove all Unship artifacts and run:
```bash
npx -y @unship/cli@latest check --json
```
Do not claim cleanup is complete until the check reports clean.
## Best Practices
- Keep Unship work local and temporary.
- Preserve the existing app design language unless the user explicitly asks for a different direction.
- Avoid unrelated refactors while variants are temporary.
- Do not put custom tabs, app preferences, or permanent switchers into product UI for Unship comparisons.
- Keep inactive options safe: avoid duplicate active IDs, submit controls, global scripts, analytics triggers, focus traps, destructive side effects, and stateful providers.
## Limitations
- Unship does not decide which variant wins; the human chooses.
- Unship does not replace design review, browser QA, accessibility checks, or production release validation.
- Unship is not intended for production traffic, remote analytics, or persistent product experiments.
## Security & Safety Notes
- Run commands only in a local project the user has authorized you to modify.
- Treat generated variants as temporary code that must be cleaned before release.
- Before destructive cleanup, confirm the selected option label when the user's choice is ambiguous.
- If a baseline build or typecheck already fails before Unship edits, report that baseline state and keep variant work isolated.
## Common Pitfalls
- **Problem:** Hidden variants override `hidden` with CSS.
**Solution:** Preserve `[hidden] { display: none !important; }` near variant-specific CSS when needed.
- **Problem:** The user says "keep the second one" after more changes.
**Solution:** Confirm the exact group and option label before editing source.
- **Problem:** The comparison grows into a broad redesign.
**Solution:** Reduce scope to the smallest section, state, or flow that can be judged in the running app.
## Related Skills
- `@webapp-testing` - Use for browser-based functional checks after frontend changes.
- `@mobile-design` - Use when comparing mobile-specific UI patterns and platform constraints.Related Skills
zustand-store-ts
Create Zustand stores following established patterns with proper TypeScript types and middleware.
zoom-automation
Automate Zoom meeting creation, management, recordings, webinars, and participant tracking via Rube MCP (Composio). Always search tools first for current schemas.
zoho-crm-automation
Automate Zoho CRM tasks via Rube MCP (Composio): create/update records, search contacts, manage leads, and convert leads. Always search tools first for current schemas.
zod-validation-expert
Expert in Zod — TypeScript-first schema validation. Covers parsing, custom errors, refinements, type inference, and integration with React Hook Form, Next.js, and tRPC.
zipai-optimizer
Ultra-dense token optimizer skill for prompt caching, log pruning, AST-based inspection, and minified JSON payloads.
zeroize-audit
Detects missing zeroization of sensitive data in source code and identifies zeroization removed by compiler optimizations, with assembly-level analysis, and control-flow verification. Use for auditing C/C++/Rust code handling secrets, keys, passwords, or other sensitive data.
zendesk-automation
Automate Zendesk tasks via Rube MCP (Composio): tickets, users, organizations, replies. Always search tools first for current schemas.
zapier-make-patterns
No-code automation democratizes workflow building. Zapier and Make (formerly Integromat) let non-developers automate business processes without writing code. But no-code doesn't mean no-complexity - these platforms have their own patterns, pitfalls, and breaking points.
youtube-summarizer
Extract transcripts from YouTube videos and generate comprehensive, detailed summaries using intelligent analysis frameworks
youtube-full
Fetch YouTube transcripts, search videos, browse channels, and extract playlists via TranscriptAPI — no yt-dlp, no Google API key, works from any cloud server.
youtube-automation
Automate YouTube tasks via Rube MCP (Composio): upload videos, manage playlists, search content, get analytics, and handle comments. Always search tools first for current schemas.
yield-intelligence
Passive income portfolio analysis — activate when user asks about dividend yields, Treasury rates, REIT income, monthly passive income goals, or portfolio yield optimization. Scans 4 asset classes, ranks by risk-adjusted return, and builds allocations targeting a specific monthly income.