Generate a set of prototypes into a single interface with a set of controls making enabling preview options.
- Related data sources
- Number of variants to generate (default 5)
Setup:
- Gather Context: as much context about the task at hand.
- Extract as much detail as possible from the source component/ui (features, interface, actions, buttons, badges)
- What is the business goal?
- What tools, UI libraries, css strategy are we using?
- What are the UX patterns for the project?
- What is entities are being used? are there common type definitions?
- Collect project reference types, data patterns, and relevant entities for the task.
- Enrich Context: trusted internet sources for: UX best practices, common solutions applicable to this task.
- Generate Data:
- Common json data set to be used by all prototypes for presentation.
- Should be in .ts file with any helper functions required to assist presentation such as filter or format functions.
- Note fields already available and separate fields that are additions or proposed
- Align for excelence: Consider prompting UI or UI agents to help build prompts or explore the web for UI prompting strategies for improving our outcome. Make the results impressive.
Build:
- Find a place to put the prototype close to existing work but not overriding any existing work
- Create a new route page:
- With controls for various data states
- Ability to show or hide each prototype component
- Placeholders for and each resulting component
- Generate Variants: num variants, each with different prompot focusing in a different strategy or angle while addressing core requirements
- Separate sub agents with different prompts with guidance for how to improve the UI.
- Each agent should produce its own component.
- When all agents complete combine all prototype components into a page so that they can all be seen togeather and evaluated.
Review:
- Ensure pototypes pass type checks
- Have specialized subagent review UI in browser via MCP and confirm that presentation meets expectations (retry as many times as neeted until UI agent is satisfied)
Return:
- Url for preview
- Relevant data sources (web sites, images, project source or docs...)
- Key features
- Variant list with approach and prescribed use case
IMPORTANT:
- Only accept the highst teir of quality, anything less is unacceptable
- Perform all work using sub agents, concurrently where possible