Design System Proving Ground
Can your engineer find THE form standard in 30 seconds?
A split screen. Left: a design system homepage with 404 errors, raw HTML patterns, and three competing "Forms" sections. Right: the same site with a single "Forms" entry, a standard process chain, and a meeting agenda form built from existing atoms. The gap between museum and proving ground.
Problem
A problem well-stated is a problem 80% solved
Situation: An engineer arrives at stackmat.es to build a form. Forms appear in 3 sidebar locations. The /forms CTA is dead. CRM pattern pages use raw HTML (L:R 1:15). "Coming Soon" stubs ship as complete. The form atoms are healthy (L:R 1.4:1, TanStack integration solid, accessibility thorough) — but the standard process that connects them is invisible.
Intention: One unified forms section with THE standard process (Zod -> Server Action -> FormShell -> Name Attrs -> TanStack), a shipping checklist, lib-only CRM patterns, and a meeting agenda form that proves composition handles real domain complexity.
Obstacle: Only 1 organism (FormShell) exists — composition depth is unproven. CRM pages violate the Prime Directive they're supposed to showcase. Navigation structure evolved organically into 3 competing locations.
Hardest Thing: Proving that the lib can compose real domain forms without needing local components. If the meeting agenda form requires a _components/ directory, the lib has fundamental gaps that no amount of documentation fixes.
Priorities
- What's broken that everyone sees? Forms 404, CRM raw HTML, stubs as content
- What's the standard nobody can find? Zod -> Server Action -> FormShell -> Name Attrs -> TanStack
- What proves the system works? CRM rewrite at L:R 1:3
- What stress-tests the system? Meeting agenda form from existing atoms only
- What prevents regression? 10-point shipping checklist
Progress
- Pictures — Outcome + dependency maps
- PRD Spec — Intent, stories, build, screen, navigation contracts
- Prompt Deck — 5-card pitch
Scorecard
Priority Score: 576 (Pain 4 x Demand 3 x Edge 4 x Trend 4 x Conversion 3)
| # | Priority (should we?) | Preparedness (can we?) |
|---|---|---|
| 1 | Pain: 4 — dogfood found 10 issues. /forms 404. CRM L:R 1:15. 3 nav locations. | Principles: 4 — Prime Directive clear. Three-mode architecture defined (IMPORT -> COPY -> USE). |
| 2 | Demand: 3 — internal builders use forms daily. Pattern seekers have shadcn. No external signal. | Performance: 2 — form atoms pass (L:R 1.4:1). No composition metrics. No shipping gate. |
| 3 | Edge: 4 — three-mode architecture novel. Anti-pattern docs rare. Meeting form as stress test unique. | Platform: 3 — atoms built, TanStack solid, 1 organism. CRM patterns broken. |
| 4 | Trend: 4 — design systems mainstream. AI code gen needs machine-readable standards. | Protocols: 2 — no standard process documented. No shipping checklist. |
| 5 | Conversion: 3 — ~60% built. Atoms healthy. Missing nav, standard process, CRM rewrite. | Players: 2 — internal only. No external users or contributors. |
| Metric | Target | Now |
|---|---|---|
| Form nav locations | 1 | 3 |
| CRM pattern L:R ratio | 1:3 | 1:15 |
| Organisms in lib | 3+ | 1 |
Kill signal: If teams don't use meeting instrument within 30 days of ship, demote to Explorations. If MeetingAgenda requires >3 new lib atoms, stop and fill lib first. Kill date: 2026-06-30.
Context
- Meetings PRD — Defines the meeting workflow this PRD stress-tests form patterns for
- Prompt Deck PRD — Design system already showcases PromptDeck, complementary
- Sales CRM PRD — CRM pattern rewrite (DS-04) improves CRM form showcase
- Standards — Where proven form patterns graduate to
- Smart Contracts — Standard fittings thesis: the form chain IS a standard fitting
Questions
What's the real cost of a design system that showcases violations of its own rules?
- If the CRM patterns bypass every lib atom, what message does that send to the next engineer who visits?
- Is the meeting agenda form a genuine stress test or a rigged demo — what domain complexity would actually break the system?
- At what point does "proving ground" become "proving what we already believe" instead of "discovering what fails"?