Prototyping workflow
The end-to-end flow from a raw problem to a shareable prototype — Ramble Refiner, build vs plan mode, tests, and collaboration.
Kickoff — from ramble to prompt
Every project starts by listening. A sponsor or stakeholder describes a problem — often messy, sometimes half-formed. Don't try to structure it in the conversation. Capture it raw.
Take the raw notes into ChatGPT and run the Ramble Refiner prompt. Its job is to:
- Refine the idea into something coherent
- Surface what the sponsor actually means vs. what they said
- Produce a first-draft prototype prompt you can paste into Lovable
Build loop — pick the right mode
Once the prototype is up, changes come in three sizes. Pick the mode that matches the size, not the one that feels fastest.
documents/features/ become the audit trail of how the prototype evolved — useful for handoff, retros, and onboarding the next Accelerator.Testing — free by default, deeper on demand
Plans generated in Plan mode include tests by default. That's usually enough for a prototype. When you need more coverage — a critical flow, a regression you keep hitting, a feature going to real users — run a dedicated test plan.
- Create a plan whose only goal is to generate tests
- Let it write the tests, run them, and produce a coverage matrix
- Save the resulting test plan to
documents/tests/
Collaboration — two paths
Who's giving feedback determines how you share it.
Artifacts checklist
By the end of a prototyping cycle, these should exist:
- Refined prompt (in the project or the Ramble Refiner chat)
- Feature plans in
documents/features/ - Test plans in
documents/tests/ - Live prototype URL (Lovable or markup.io depending on audience)
- A short list of what you learned — the whole point of the exercise
