About plan-prototype
Plan Prototype is a skill for answering one planning question with a small, disposable experiment. It helps developers validate UI or logic decisions before implementation, then report evidence, uncertainty, and design implications.
Key features
- 01Frames an experiment around one planning question
- 02Chooses visual or logic prototypes as appropriate
- 03Prefers existing tools and in-memory state
- 04Marks prototype code as throwaway
- 05Reports observations, conclusions, and design impact
Use cases
- 01Validate a UI layout or flow
- 02Explore state, data, or API behavior
- 03Compare alternative approaches
- 04Gather experiment evidence for a plan or design document
Install
Copy the command and run it in a terminal.
npx skills add https:/ / github.com/ klausborges/ workflow-skills --skill plan-prototypeAfter installing, type/plan-prototypein Claude Code to confirm it loaded.
Bundled files
References and scripts installed along with SKILL.md.
skill.tomlagents/openai.yamlreferences/workflow-language.mdDetails
From the project READMEPlan Prototype
Answer one planning question with the smallest disposable experiment.
Use workflow-language.md when the result feeds a Plan or durable doc.
Scope
- State the question and the evidence that would answer it.
- Choose a visual prototype for UI, diagrams, flows, or option canvases; choose a logic prototype for state, data, API, or business-rule questions.
- Ask about mode only when the question does not make it clear.
- Compare multiple variants only when comparison is the question. Do not add a switcher, persistence, polish, abstractions, or tests unless they materially improve the answer.
- Prefer existing project tooling and in-memory state. Mark prototype code as throwaway.
Put a visual prototype in the real page/flow only when real density, navigation, or data is necessary to answer the question. Keep logic behind a small interface only when portability is itself useful to the experiment.
Closeout
Report the answer, supporting observations, uncertainty, and recommended design consequence. Record it in an existing Plan, doc, or ADR only when that owning artifact exists and the conclusion belongs there; chat may be sufficient.
Delete or absorb only prototype artifacts created by this task after their disposition is clear. Preserve preexisting prototypes and unrelated drafts. Do not turn a useful experiment into production implementation without explicit authorization.