monthly-review

Monthly Review — User-Driven

Your job: aggregate weekly data, show it raw, ask questions, format the user's answers. NOT your job: analyze project/area alignment, generate insights, recommend project changes, edit about-me.md.

---

Step 1: Determine the month

Identify the month being reviewed (current or just-completed).

Step 2: Aggregate raw data (mechanical only)

Pull these. No interpretation.

1. Weekly grades — from each weekly review section in Periodic_notes/YYYY/weekly/YYYY-W##.md (4–5 weeks)
2. Monthly habit rates — read habit completion from daily-note frontmatter for the month, sum/average per habit
3. Tasks — from the monthly note: count [x] vs [ ] for "Must Complete" and "Should Progress"
4. Project status changes — read 01_Project/*.md frontmatter, list any project where project-status changed this month
5. Headlines from each week — pull the user's verbatim "headline" or top section from each weekly review

Step 3: Show the data table to the user

## [Month YYYY] — Raw Data

Weekly grades:
| Week | Grade | User's headline (verbatim) |
| W##  | A-    | "shipped feature X"         |
| W##  | B     | "low capacity, family week" |
| W##  | ?     | [no review done]            |
| W##  | B+    | "catch-up week"             |

Habits (month average):
- [Habit]: X/N days (X%)
- [Habit]: X/N days (X%)

Monthly tasks:
- Must Complete: X/Y done
- Should Progress: X/Y done

Project status changes this month:
- Project A: Doing → Done
- Project B: Planned → Doing
- (none) for Project C

Daily journaling: X/N days had entries

No commentary. No "trends". No emojis beyond what's in source data.

Step 4: Ask 4–6 sharp questions, ONE AT A TIME

Generate questions from the data. Wait for each answer.

Default question set (adapt based on data)

1. Headline question — "Looking at the data, what's the headline of this month in one sentence?"
2. Real progress — "Which project or area got real progress this month? (Not 'I worked on it' — actual deliverable shipped.)"
3. Stalled callout — "Project X had no status change this month and was on the plan. What actually happened?" (use real project name from data)
4. Pattern question — "Habit Y was Z% — same as last month. Same blocker or different?"
5. Decision — "Continue, pause, or drop Project X? Pick one."
6. Carry-forward — "One thing to start next month, one thing to stop?"

Question rules

Same as weekly-review:

• ✅ Use real data in the question (specific project names, specific numbers)
• ✅ Force decisions, not exploration
• ❌ No therapy-speak, no leading questions, no sympathy
• ❌ Don't fish for answers AI already has in mind

Step 5: Accept answers verbatim

User answers via voice or short text. Their words go directly into the review. Do NOT rephrase.

Light formatting only: capitalize, fix typos, add bullets. Nothing else.

Step 6: Output format

Append to the monthly note under ## Monthly Review:

## Monthly Review — [Month YYYY]

### Headline
[User's answer to Q1, verbatim]

### Data
- Weekly grades: [grades]
- Habits (month avg): [rates]
- Tasks: X/Y must, X/Y should
- Projects changed: [list]

### Real progress
[User's answer to Q2, verbatim]

### Stalled
[User's answer to Q3, verbatim]

### Pattern
[User's answer to Q4, verbatim]

### Decisions
- [User's project decisions from Q5]

### Start / Stop next month
- Start: [user's answer]
- Stop: [user's answer]

Section headers reflect what the user said, not pre-canned templates.

Step 7: long-term-context update — explicit gate only

NEVER auto-edit 99_metadata/about-me.md (or whatever long-term self-profile the user maintains).

After the review is written, ask ONCE:

"Did anything this month change a long-term pattern in about-me.md? (Y/N)"

• N → done. No edit.
• Y → ask: "What changed? In which section?"
  • User describes the change in their own words
  • You draft a diff (small, surgical — touch only the section the user named)
  • Show the diff
  • Ask: "Apply? (Y/N)"
  • Apply only on explicit Y
  • Update last-updated in frontmatter

Never propose pattern changes the user didn't name. about-me.md is the user's self-profile, not yours.

---

What this skill does NOT do

• ❌ Generate "What Went Well / What Went Poorly"
• ❌ Generate "Actionable Insights" or recommendations
• ❌ Analyze project/area alignment for the user
• ❌ Suggest projects to drop, pause, or pivot
• ❌ Identify patterns the user didn't name
• ❌ Auto-edit about-me.md
• ❌ Grade the month
• ❌ Compare to previous months unless user asks
• ❌ Soften or "improve" the user's wording

Reminder for the AI running this skill

A monthly review's value is the user processing the month, not reading a summary of it. If you analyze for them, you steal the value. Aggregate data → ask sharp questions → format their answers. That's the entire job.