G3 — Structured Weekly Review ritual

[ColdByDefault]

Overview

A guided weekly review flow that the assistant pre-populates from the user's actual data. Output is stored as a dated snapshot. Based on GTD and 12 Week Year review practices — but the AI does the data gathering automatically.

Flow (5 steps)

1. What shipped? — closed tasks this week (auto-loaded)
2. What slipped? — tasks moved past due date (auto-loaded)
3. What decisions did I make? — decisions from this week (auto-loaded)
4. What am I committing to next week? — user fills in (LLM suggests from open items)
5. One thing to reflect on — free-form note

New model

model WeeklyReview {
  id          String   @id @default(cuid())
  userId      String
  isoWeek     String   // "2026-W17"
  shipped     Json
  slipped     Json
  decisions   Json
  commitments String
  reflection  String?
  createdAt   DateTime @default(now())
  @@unique([userId, isoWeek])
}

Files

• prisma/schema.prisma + migration
• lib/weekly-review/
• app/(app)/reports/weekly-review/page.tsx or new route
• components/reports/WeeklyReview.tsx
• messages/de.json + messages/en.json

Acceptance Criteria

• One-click start launches the review, pre-populated with real data
• Each section is editable before saving
• Saved reviews are stored and viewable historically
• LLM can generate a natural summary of the completed review

[ColdByDefault]

Soft dependencies on: #84 (E1 — Decision Outcome Journal) and #89 (G1 — Delegate Tracker)

G3 works without either — it uses existing tasks, meetings, and decisions data. However:

• If E1 is merged, the review can include outcome ratings on this week's decisions
• If G1 is merged, the review can include delegated tasks that slipped

No hard blocker, but richer output if both are done first.