Structural Resilience Report
This is a real report. We ran our audit instrument against our own design system and published the result, scores, gaps and all. Your report follows the same structure, built from your system instead of ours.
Six dimensions, no single grade.
A design system is never one number. The radar shows where the system is strong and where the next quarter of work belongs. Balanced shapes need care; spiked shapes need specific investment. Ours spikes on tokens and accessibility, and sags where one operator does by hand what tooling should hold.
Structured token config with provenance fields and a 124-token validation registry. Held back by fields still marked provisional pending designer confirmation.
Production components validate against a nine-rule engine. No complete inventory yet beyond the validated set.
Append-only decision log and documented propagation rules. Governance is real but rests on a single operator.
Tokens are canonical in code. The Figma round trip is manual, so drift between design files and production is caught by review, not tooling.
Contrast verified pair by pair, with boundary cases documented rather than hidden. One accent passes only at large sizes and the system says so.
This dimension reads how a team relates to the system and to each other. A single-operator system has no team, so there is nothing here to score. On team-scale engagements, this is where the trust and adoption findings live.
Three findings, in the order we would fix them.
- HIGHF-01
An accent that passes only at large sizes
Electric Crimson (#FF3B30) is the primary accent and clears contrast only for large text on light backgrounds. Nothing in the build pipeline stops a future surface from setting it at body size.
Why it matters: The most-used accent is the most likely to drift into an illegible state. The constraint is documented; it is not yet enforced.
Next step: Encode the pair rule as a blocking check so misuse fails the build instead of shipping.
- MEDF-02
A provisional token without a designer signature
One text token carries a provisional value, pending a designer confirmation gate. It renders with an annotation, which is honest, but the gate has been open since March.
Why it matters: A provisional value that ships long enough becomes a de facto decision nobody made.
Next step: Schedule the confirmation review, or promote the value with a dated decision.
- MEDF-03
The sync gap between config and canvas
The token config is the source of truth and the code follows it. Design files follow it by hand.
Why it matters: Manual sync holds at one operator. It fails quietly the first week two people edit in parallel.
Next step: Wire the variables pull so the canvas reads the same source the site does.
The instrument reads the system. The audit reads the room.
Everything above came from machine-readable surfaces: tokens, components, contrast pairs, build output. The full engagement covers what no tool reaches. Whether the system is funded and owned. Whether your team trusts it enough to stop working around it. Which finding to fix first, and which to leave alone. That ordering is judgment, and it is the part you are paying for.
Prepared July 2026 · A real audit of our own system · heyhi.design