AGENT SKILL

prototype

Prototype a state model, business logic, or UI with throwaway code that answers a design question.

License 许可
MIT
所属仓库
BetoFrega/skills
最近更新
2026-11-14
Stars
★ 0
Forks
0

01 前言 Front Matter

name
prototype
description
Prototype a state model, business logic, or UI with throwaway code that answers a design question.

02 资源文件 4

03 SKILL.md 正文

Prototype

Identify the question and completion criterion. For product questions, read canonical contracts/evidence through consolidate, including external configuration. A prototype that works sets no product rule and authorizes no release.

Pick a branch

  • Logic or state model: LOGIC.md, one shareable HTML file with free-play actions and tabbed walkthroughs, usable by a non-developer.
  • Appearance: UI.md, distinct variants on one route, selectable through a URL parameter and floating bottom bar.

Infer from prompt/code or ask when ambiguous. If the user is unreachable, prefer logic for backend modules and UI for pages/components; state the assumption at the prototype's top.

Rules that apply to both

  1. Throwaway: place code beside its intended module/page, named as a prototype. Temporary UI routes follow existing project routing conventions.
  2. Runnable: UI starts with one project-task-runner command; logic opens as one HTML file.
  3. In-memory: if persistence is the question, use scratch database/file storage clearly named "PROTOTYPE, wipe me".
  4. Minimal: no tests, abstractions, or error handling beyond runnable behavior.
  5. Visible state: show full relevant state after each logic action or UI switch.
  6. Capture: integrate validated decisions in real code. Commit the complete rerunnable prototype as a primary source on a throwaway branch outside main; link that branch from the implementation issue. Record question/verdict in the issue or commit. Main retains only validated decisions.

Return supported learning and accepted choices through consolidate within existing documentation authority, preserving unresolved proposals/evidence limits. Use next steps for completion, meaningful learning review, or ready continuation; feature delivery and follow-up work retain their own scope.

04 同仓库更多 Skills