Weekly review

Weekly Review

Closes the week with a four-part review — wins, carryover, metrics, next week's top 3 — and appends it to a running weekly-reviews.md so the log compounds into a real track record. The standard: evidence over vibes (wins name artifacts, metrics have numbers or an honest "not tracked"), and the log file is append-only — never rewritten.

When to Use

• Scheduled Friday run (the blueprint fires at 17:00 Fri with "Run my weekly review").
• "Wrap up my week", "weekly review", "what actually got done".
• Not for: the daily morning kickoff (use morning-brief) or extracting actions from one meeting (use meeting-notes-actions).

Quick Reference

Action	Command / Call
Week bounds	date -v-monday "+%Y-%m-%d" (macOS) / date -d "last monday" +%F (Linux)
Find the log	find ~/Documents -maxdepth 3 -name "weekly-reviews.md" 2>/dev/null
Read last entry	tail -60 "weekly-reviews.md"
Week's file evidence	find <projects-dir> -newermt "$(date -d 'last monday' +%F)" -type f 2>/dev/null | head -20
Append entry	cat >> "weekly-reviews.md" (append only — never > overwrite)

Procedure

1. Precheck — locate weekly-reviews.md (find, or the path remembered from prior runs). Missing? Ask where to create it (default ~/Documents/weekly-reviews.md), then start it with a # Weekly Reviews header. Compute the week's Mon–Fri dates with date for the entry header.
2. Gather last week's promises — tail -60 the log and pull the previous entry's "Next week's top 3." These are this week's scorecard.
3. Collect evidence — from what's available: daily notes, files modified this week (find -newermt), calendar events, anything the user tells you. Then ask the user 3 quick questions: biggest win? what stalled? any metric moved? Their answers outrank inference.
4. Draft the entry — Wins (concrete, artifact-named); Carryover (unfinished items + one-word reason: blocked/deprioritized/underestimated); Metrics (number, delta vs. last entry if logged, or "not tracked"); Next week's top 3 (specific, finishable, at most 3). Score last week's top 3 as done / partial / dropped.
5. Confirm — show the full drafted entry. On explicit approval, proceed; edits first if requested.
6. Append — cat >> weekly-reviews.md with the entry preceded by \n---\n. Append-only: never open the file with >, never edit past entries. Confirm with tail -5 that the new entry is at the bottom.

Output Template

---
## Week of {Mon YYYY-MM-DD} → {Fri YYYY-MM-DD}

**Last week's top 3 — scorecard:** {done: n}/{3} ({item: done/partial/dropped}, ...)

### Wins
- {win — artifact/evidence}

### Carryover
- {unfinished item} — {blocked | deprioritized | underestimated}

### Metrics
- {metric}: {value} ({delta vs last week | "not tracked"})

### Next week's top 3
1. {specific, finishable}
2. ...
3. ...

*Logged {YYYY-MM-DD HH:MM}*

Pitfalls

• Log file overwritten instead of appended — a > where >> belonged deletes months of history. Always >>; verify afterward with tail -5 plus a line count sanity check (wc -l should grow, never shrink). If history was lost, check for editor backups/Time Machine immediately and tell the user before doing anything else.
• Wins are vague vibes ("made good progress") — every win names an artifact: a file, a shipped feature, a sent proposal, a number. If no artifact exists, it goes to Carryover, not Wins.
• No previous entry exists, so the scorecard breaks — first run is legitimate: write "First entry — no scorecard" and set the top 3 so next week's review has something to score.
• Scheduled run fires with the user absent and no answers to the 3 questions — degrade gracefully: build the entry from file/calendar/notes evidence, mark it (auto-draft — confirm wins with me), and still append so the chain never breaks. Don't fabricate answers on the user's behalf.
• Top 3 becomes top 7 — the cap is the discipline. Three items maximum; everything else is a backlog note inside Carryover.

Verification

• Entry matches the template — all four sections plus scorecard present
• Entry was appended (tail -5 shows it at the bottom; wc -l grew; earlier entries untouched)
• Every win cites an artifact; every metric has a number or an explicit "not tracked"
• Last week's top 3 were scored, or "First entry" is stated
• Exactly 3 (or fewer) next-week priorities
• User approved the entry before append (or auto-draft is labeled as such)