Quality framework and token checklist
Two working tools I designed and built with AI, over one scan of the codebase: an inventory of every colour, spacing and size in use, and a worklist for deciding what each should become.
About the tool
Tool · Built with AI · Design systems- Why
- Tokens had to be clean before AI could build screens that match the design system. Nobody knew how many colours, spacings and sizes were really in use.
- Built by me
- I designed and built both tools myself, using AI (Claude Code) to write the code, run the scan across the codebase and sort the results. No developer time was needed to build them.
- The scan
- One scan across two repositories: 4,881 files and 15,569 colour declarations. Every tool reads the same file, so they can't disagree.
- Colour audit
- The raw inventory: the full palette by family, every declaration, and the breakdown by component and CSS class.
- Colour consolidation
- Every distinct colour mapped to a proposed scale, sorted into adopt, replace, review and decide, with a decision you can record and export.
- Dimensions
- Rounded corners, border widths, spacing and sizing measured against proposed scales, with the same four decisions plus report only.
- Checks
- Visual-fidelity gates, token-coverage checks and validation of generated code, with fixes pushed back into Figma.
- Names
- Product and repository names have been replaced with neutral ones.
Screens
Get in touch
Looking for a product and UX/UI designer for a contract role? I'd like to hear about it.
Start a conversation- Connect
LinkedIn