UX Audit — Expertise domain detail

Layers: L1 ✅ · L2 desktop dark/zero-run/mature ✅ · L3 drill-down ✅.

Patterns in use

Pattern	Rating	Evidence
Center Stage	⚠️	Fit statement/chart dominate; action-relevant gaps are secondary.
Titled Sections	✅	Trend, metrics, coverage, Top 5 and anchoring are separated.
Overview + Detail	✅	Top 5 drills into all rules and individual rules.
Failure + Retry	⚠️	Domain fetch is covered; independent lessons fetch is not.
Forward Momentum	—	No-practice and anomaly states explain but offer no action.

✅ 亮点 — Unreliable fits do not invent a number. The page explains why projection is withheld (Detail/index.tsx:263+).
✅ 亮点 — Zero practice is not failure. A purpose-built state replaces a meaningless chart (:135-149).
✅ 亮点 — Canonical gaps stay visible. Empty layers are not filtered out (CoverageCloud.tsx).

🔴 Lessons failure masquerades as zero lessons. Only domain error/loading is read; lessons become lessons ?? [] (Detail/index.tsx:77-78, :224-230).
🟠 No-practice is a dead end. No start matching Topic, inspect likely matches, or edit boundary action.
🟠 Detected gaps have no corrective action. Empty layers, unused and unanchored lessons are display-only.
🟠 Expertise lifecycle is read-only. No edit, pause, archive or delete exists.
🟡 Implementation metrics precede the decision. “What needs attention / what next” has no first-class block.