Weekly Skill

What this does

Builds the weekly review note at 50-daily/reviews/weekly/{YYYY-Www}.md for the current ISO week, synthesizing it from that week's daily notes + _meta/orient.md, then runs one consolidated interview to confirm next week's plan. The reflective sibling of /daily.

Determinism split (read this first): the mechanical steps below are set — run them exactly, no improvising, no extra exploratory calls. The steps marked (judgment) are where thinking is wanted: synthesizing themes, summarizing project movement, and suggesting next week's Top 3. Don't fabricate in the judgment steps — reason over what the dailies actually say.

Mode detection (mechanical)

1. Target week. If the user passed an arg (YYYY-Www, or a date inside the week), use that week. Otherwise use the ISO week containing today.
2. Compute the week (ISO 8601): weeks run Monday → Sunday; a week belongs to the year of its Thursday. Derive: the Monday (start) date, the Sunday (end) date, the YYYY-Www label, and a human range like June 8–14, 2026.
   • Worked example: today 2026-06-14 (Sunday) → 2026-W24, Mon 2026-06-08 → Sun 2026-06-14, range "June 8–14, 2026".
3. Path: 50-daily/reviews/weekly/{YYYY-Www}.md. Then branch on its state:

Setup (BUILD) — mechanical

1. Read the week's daily notes. For each date Monday..Sunday of the target week, read 50-daily/{YYYY}/{MM}/{YYYY-MM-DD}.md if it exists. From each, pull: Top 3, Wins, Notes / log, Carry forward to tomorrow, and any - [x] done / - [ ] open items. (Missing days are normal — note which were absent; don't invent their content.)
2. Read context: _meta/orient.md (Active projects + This week's focus) and the most recent prior weekly review in 50-daily/reviews/weekly/ (for continuity — you'll link it).
3. Create the note from templates/weekly-review.md, substituting {{date}} → the Sunday (or today if the week isn't over), {{iso-week}} → YYYY-Www, {{date-range}} → the human range. Create parent folders if missing. Add an info callout linking the previous week, e.g. > Follows [[2026-W23]]. If any in-range day was missing and you reconstructed loosely, add backfilled: <today> to frontmatter and say so in the callout.

Draft the sections (judgment)

From the material gathered in Setup, draft each template section — reason over the dailies, cite real items, invent nothing:

• What got done — completed items + Wins across the week's dailies.
• What didn't (and why?) — Top-3 / carry-forward items that persisted unfinished, with the brief reason the dailies give (or "no reason logged").
• Themes this week — 2–4 patterns that recurred across days.
• Progress on active projects — per project in orient.md / 10-projects/, summarize the movement visible in the dailies (or "quiet — no daily activity").
• Wins to celebrate — the headline wins.
• Honest reflection (good / bad / habits) — draft a candid read: 2–3 genuine bright spots, 2–3 real misses or bad patterns, and the behavioral habits to fix / must-dos for next week. Ground every point in a dated daily-note observation. Be honest, not flattering — name the patterns (perpetually-carried backlog, conditional "if time" Top-3 slots, days that never got closed, declared-then-abandoned threads). This is the part the user most wants challenged; don't soften it into praise.

Interview (judgment) — two beats

Beat 1 — Plan. Present a tight summary of the draft, then ask in a single message:

1. "Anything wrong or missing in what got done / didn't / themes?"
2. "Top 3 for next week?" — offer a suggested draft built from orient.md's focus, unfinished carry-forwards, and any live deadlines.
3. "Anything to deprioritize next week?" Wait for the response, then fill What got done / didn't / themes / project progress / wins (corrected) and Top 3 for next week + To deprioritize.

Beat 2 — Reflect (run this every week — it's the heart of the ritual, not optional). First share your own honest read (from the reflection draft above): a couple things that were genuinely good, a couple that were bad, and the habits to fix / must-dos for next week — candid, dated, not flattering. Then ask the user for theirs, in a single message:

• "What do you think was good and bad this week?"
• "Any bad habits that cost you — and anything to make sure you do next week?" Wait for the response, then write the Honest reflection section, synthesizing both your read and the user's into a cleaned-up form. The user's own words lead; your read supports. (Priorities/deprioritizations are already handled in Beat 1 — if the reflection changes them, update those sections too.)

Wrap-up (mechanical + confirm)

1. Fill Open questions surfaced from unresolved threads in the week's dailies. Offer to append them to _meta/questions.md (only on confirmation).
2. Fill Notes for orient.md update with concrete changes the week implies. Do NOT edit orient.md yourself — per house rules, propose the changes as a diff and wait for the user to apply them.
3. Leave the note status: active. Report the path and the proposed orient.md diff.

Rules

• Never overwrite a filled weekly note. The Wins/Top-3 check in Mode detection short-circuits to DONE first (same guard as /daily-catchup).
• Don't invent. Every "got done / didn't" line traces to a daily note. Sparse week → say so and reconstruct conservatively.
• Mechanical steps are fixed (week math, paths, which files to read, file creation) — no improvising, no extra calls. Judgment steps may think (synthesis, suggestions).
• No Notion calls. Weekly review is reflection over the vault's own dailies; it does not pull a task manager. (That's /daily's job, and only when one is configured.)
• Use [[wikilinks]] for projects, people, and concepts that have vault notes.
• _meta/orient.md, _meta/profile.md, _meta/memory.md are propose-as-diff only.