Skip to content
Agentmart
Submit a tool

prototype

Use when you need to sanity-check a state model or interaction before building it, or when you need to decide what a UI should look like.

Skill by hellsan631

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/hellsan631-prototype

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use when you need to sanity-check a state model or interaction before building it, or when you need to decide what a UI should look like. Builds a throwaway prototype to answer the question.

Use cases: design

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/prototype

What's inside

Skill name
prototype
Description
Use when you need to sanity-check a state model or interaction before building it, or when you need to decide what a UI should look like. Builds a throwaway prototype to answer the question.
License
MIT
Instructions
23 lines in SKILL.md · view on GitHub

Files (4)

  • LOGIC.md
  • SKILL.md
  • UI.md
  • openai.yaml · agents/

SKILL.md (first lines)

# Prototype

A prototype is **throwaway code that answers a question**. The question decides the shape.

## Pick a branch

Use the user's prompt and the surrounding code to identify the question. If they do not settle it and the user is available, ask:

- **"Does this logic / state model feel right?"** → [LOGIC.md](LOGIC.md). Build one shareable HTML file with free-play buttons and tabbed guided walkthroughs. It must push the state machine through cases that are hard to reason about on paper, and a non-developer must be able to drive it.
- **"What should this look like?"** → [UI.md](UI.md). Generate several radically different UI variations on one route. Make them switchable through a URL search parameter and a floating bottom bar.

The branches produce different artifacts, and choosing the wrong one wastes the prototype. If the question remains ambiguous and the user is unavailable, use the surrounding code: choose logic for a backend module and UI for a page or component. State the assumption at the top of the prototype.

## Rules for both branches

1. **Mark it as throwaway code from the start.** Put the prototype next to the module or page it tests so its context is visible. Name it as a prototype so a casual reader does not mistake it for production code. For throwaway UI routes, follow the project's existing routing convention. Do not add a new top-level structure.
2. **Make it one step to run.** Start a UI prototype with one command in the project's task runner, such as `pnpm <name>`, `python <path>`, or `bun <path>`. Make a logic demo a single HTML file the user can open by double-clicking.
3. **Keep state in memory by default.** Add persistence only when the prototype is checking persistence. If the question explicitly involves a database, use a scratch database or a local file with a clear "PROTOTYPE, wipe me" name.
4. **Build only enough to answer the question.** Add no tests or abstractions, and no error handling beyond what makes the prototype runnable. The point is to learn something fast.
5. **Show the state.** After every logic action and every UI variant switch, print or render the full relevant state so the user can see what changed.
6. **Capture the result.** Fold every validated decision into the real code. Preserve the prototype as a **primary source** by committing it to a throwaway branch outside main and leaving a context pointer to that branch on the implementation issue. Record the answer, including the verdict and the question it settled, in the issue or a commit. Keep only the validated decision on the main branch.

From hellsan631/skills, MIT.

More from hellsan631/skills

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Aug 29, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from hellsan631

Cite this page

prototype skill. Agent Mart. https://agentmart.dev/skills/hellsan631-prototype (last updated 2026-10-24)