# Skill: weekly-review-focus

- **Name:** Weekly Review, Scored
- **Trigger:** "run my weekly review" · once a week, Sunday works best
- **Requires:** my Focus file and Memory files · Calendar (read) if connected, it sharpens the scoring · Level: Minimum, better at Optimum
- **Output:** Each priority scored 0-10 with the evidence named, one broken promise and one win, a drafted Focus file update shown as a diff, and at most one skill suggestion from the library index. The review is saved as a dated file (review/YYYY-Wnn.md) and the applied update gets an "approved: date" line, evidence the graduation check can find.
- **Safety:** READ, then ONE gated write: it drafts the Focus file update and waits for OK before saving. It changes nothing else.

## Steps

1. Read my Focus file. List P1, P2, P3 and the week's stated top 3.
2. Gather evidence: the past 7 days of calendar if connected, and files
   created or changed this week. If you cannot see file dates or a
   calendar, skip that and ask me to name what I worked on instead. Then
   ask me exactly one question: "Anything this week the files do not
   show?" Wait for the answer.
3. Score each priority 0 to 10. The score follows evidence, not mood:
   hours actually spent, the measure actually moving. State the evidence
   in one line per priority. Do not flatter me for a half-done week; a 4
   with honest evidence beats a 7 with none.
4. Name one broken promise and one win, each with a date.
5. Check the "not now" list for leakage: did anything parked eat hours?
   If yes, say which and how many.
6. Draft the updated Focus file: this week's theme, next week's top 3,
   insights section trimmed to a maximum of 4 entries none older than
   14 days. Show the change as a diff, old line then new line, and wait
   for OK before saving. After the OK: save the Focus update, write
   "approved: {date}" at the top of the review, and save the review
   itself as review/YYYY-Wnn.md (create the folder on first run).
7. After the OK, open skills/_INDEX.md. Compare the week's biggest
   friction against the skill list. If exactly one skill would have saved
   me 30 minutes or more, offer it in one sentence: "You lost time on X.
   The {name} skill covers that, want it active this week?" If nothing
   clearly fits, offer nothing. Never offer two.

## Example run

Me: run my weekly review

AI: One question first: anything this week the files do not show?

Me: I spent Tuesday evening on the workshop outline, never wrote it down.

AI:
WEEK REVIEW · 6 to 12 Jul

P1 Client pipeline: 6/10. Two proposals out (Tue, Fri), zero follow-up
calls booked. The measure was "2 calls booked", it did not move.
P2 Health: 4/10. Calendar shows 1 of 3 planned runs. Tue run cancelled
for the workshop outline.
P3 Website: 8/10. New case study drafted and published Thursday.

Broken promise: follow-up call with Marta, promised Wed, still not booked.
Win: case study live on 10 Jul, first one in 3 months.
Not-now leakage: the podcast idea took roughly 2 hours on Thursday.

Focus file update (diff): [old top 3 -> new top 3, theme line, 1 insight
added, 1 insight expired] ... OK to save?

Me: OK

AI: Saved. One more thing: you rewrote three client emails by hand this
week. The gmail-weekly-email skill covers that, want it active this week?

## Make it yours

- Change the friction threshold in step 7: 30 minutes is the default bar
  for suggesting a skill.
- Add a spheres pass: one line per life area after the scores, if you run
  the 7-sphere or PARA memory layout.
- Move the day: Sunday evening is the default, Friday close-of-week also
  works, pick one and keep it.