moveros 4.0.8 → 4.1.0

package/src/workflows/review-week.md

@@ -1,558 +1,558 @@
----
-description: Weekly review. Clean the system, harvest wisdom, audit identity, validate strategy, implement system improvements.
----
-
-# Weekly Review (V4)
-
-**Trigger:** User runs `/review-week` (user-configured day, default Sunday).
-**Role:** Chief of Staff & Performance Psychologist.
-**Task:** Clean system, harvest wisdom, audit identity, validate strategy, reset for next week.
-**Risk Tier:** HIGH (Strategy changes require confirmation)
-
-This is the system improvement point. Issues detected during the week get fixed here.
-
----
-
-## 0. PRE-FLIGHT
-
-Global Rules are pre-loaded as CLAUDE.md / GEMINI.md. Output `[Rules Loaded - DATE]`.
-
-VAULT ROOT (resolve ONCE, use for all file paths):
-```bash
-git rev-parse --show-toplevel
-```
-Save as VAULT_ROOT. All paths below are relative to VAULT_ROOT.
-
-**Review file paths:**
-- Weekly: `{VAULT_ROOT}/02_Areas/Engine/Weekly Reviews/Week - [YYYY-WW].md`
-- Monthly: `{VAULT_ROOT}/02_Areas/Engine/Monthly Reviews/Month - [YYYY-MM].md`
-- Quarterly: `{VAULT_ROOT}/02_Areas/Engine/Quarterly Reviews/Quarter - [YYYY-QN].md`
-
-Create the folder if it doesn't exist before writing. Never write outside `02_Areas/Engine/`.
-
-LOAD (Memory Hierarchy):
-1. Auto_Learnings.md (P0) - FULL file
-2. Active_Context.md (P1)
-3. Strategy.md (P2)
-4. Identity_Prime.md (P3)
-5. Mover_Dossier.md (P3)
-6. Goals.md (if exists)
-7. Metrics_Log.md (if exists) — 30-day rolling data for strategy verdict
-8. Daily Notes from this week — read on demand in Step 2A, not all upfront (P5)
-
-REVIEW DAY: Read `review_week_day` from Active_Context.md Workflow State (default: Sunday). All "review day" references in this workflow use this user-configured day.
-
-MONTHLY MODE CHECK (date-based, run before proceeding):
-```bash
-# Is today the last occurrence of review_week_day this month?
-DAYS_LEFT=$(python3 -c "
-from datetime import date, timedelta
-today = date.today()
-next_week = today + timedelta(days=7)
-print('LAST' if next_week.month != today.month else 'NOT_LAST')
-" 2>/dev/null || node -e "
-const t=new Date(), n=new Date(t); n.setDate(t.getDate()+7);
-console.log(n.getMonth()!==t.getMonth()?'LAST':'NOT_LAST');
-")
-```
-If output is `LAST`: set MONTHLY_MODE=true. This is the user's monthly review day.
-
-QUARTERLY MODE CHECK (date-based):
-If MONTHLY_MODE=true, also check if current month is March, June, September, or December.
-If yes: set QUARTERLY_MODE=true.
-
-CONSTRAINTS:
-- All verdicts grounded in weekly data, not single-day observations
-- Strategy changes require user confirmation (HIGH-RISK)
-- Append-only: never delete strategy history
-- Sample size validation before recommending pivots
-- Label causal claims with confidence level
-- Windows: no && in commands
-
----
-
-## 1. PREREQUISITE CHECK
-
-- At least 3 Daily Notes exist for this week (minimum data for review)
-- Strategy.md exists
-- Identity_Prime.md exists
-- Today's `/analyse-day` should be run first (suggest if not)
-
-If < 3 Daily Notes: CAUTION. "Only [X] days logged. Review will be limited. Proceed anyway?"
-
----
-
-## 2. THE SYSTEM AUDIT
-
-### 2A. Read All Sources
-Using files already loaded in Pre-Flight, extract the following. Only re-read if content not yet in context:
-1. Strategy.md — Hypothesis, Metrics, Sample Sizes, Pivot Criteria
-2. Identity_Prime.md — Values, Anti-Identity, Non-negotiables, Rituals
-3. This week's Daily Notes — read one at a time, extracting: Tasks completed, Battery scores, Session Log events, Key Insights
-4. Auto_Learnings.md — full pattern database (already loaded)
-5. Goals.md — current targets (if loaded)
-6. Metrics_Log.md — last 30 days of data (if loaded)
-7. SCAN `00_Inbox/` — Quick Capture, Integrations folder
-8. SCAN `01_Projects/` — list active projects, last commit dates
-
-After each Daily Note read, output a one-line extract: `[Date]: Tasks X/Y | Energy: Z | Notable: ...`
-This keeps you from losing early-week data by the time you reach later entries.
-
-### 2B. Batch Reading Protocol
-If > 7 files to read:
-1. Split into batches of 4-5 files
-2. Read each batch, extract patterns
-3. Synthesize across batches
-4. Cross-reference with Engine files
-
----
-
-## 3. THE PROCESSING PROTOCOL
-
-### 3A. Identity Audit (The Mirror + Evolved Self Tracking)
-Compare user's behavior across ALL dailies vs Identity_Prime.md.
-- "Did we embody the identity this week, or did we slip into the Anti-Identity?"
-- If slipped: identify trigger (Lack of sleep? Fear? Overwhelm?) and prescribe fix.
-- Cite specific days: [Daily - YYYY-MM-DD] for evidence.
-- Do NOT soften the verdict. If the week was bad, say it was bad. "You planned to be [identity] this week. The data shows you were [actual behavior] on [X] out of [Y] days. That's not a rough week, that's a pattern."
-- If the same slippage appeared last week: "This is the second consecutive week of [pattern]. Once is a bad week. Twice is a habit forming. What structurally changes next week?"
-
-**Subconscious Pattern Tracking:**
-- Which evolved behaviors are becoming automatic? (Happened without conscious effort this week)
-- Which still require conscious effort? (Had to force yourself to do them)
-- Any new Anti-Identity triggers surfaced this week that aren't in Identity_Prime.md yet?
-- Track the trajectory: "Week over week, [behavior X] is becoming more natural / still requires effort."
-
-### 3B. Inbox Zero Check
-Scan ALL files in `{VAULT_ROOT}/00_Inbox/`. For each item, classify and route:
-
-| Content Type | Destination |
-|---|---|
-| Task / action item | Active project's plan.md or Active_Context Backlog |
-| Article / video / book note | `03_Library/Inputs/[Articles\|Videos\|Books]/` using template |
-| Reusable knowledge (pattern, principle, SOP) | `03_Library/` (Cheatsheets/Principles/SOPs) |
-| Person / org / place mentioned | `03_Library/Entities/[type]/` using template |
-| Decision documented | `03_Library/Entities/Decisions/` using template |
-| Project idea (not started) | `02_Areas/Engine/Someday_Maybe.md` |
-| Noise / outdated / irrelevant | Delete or move to `04_Archives/` |
-
-Process `00_Inbox/Quick_Capture.md` (if exists): parse each entry and route individually.
-Process `00_Inbox/Integrations/` folder if it exists.
-
-**Goal: Inbox must be empty by the end of review.** Every item either routed or trashed.
-
-If > 10 items: batch-present them to the user for quick triage rather than asking one by one.
-
-### 3C. Project Health & Harvest Scan
-For each active project in `01_Projects/`:
-
-**Harvest Scan (DO NOT harvest inline — just count):**
-1. Read `project_state.md` (if exists)
-2. Count Solutions Ledger entries NOT in `## ARCHIVED KNOWLEDGE (Harvested)`
-3. If unharvested entries > 0: record project name + count
-4. After scanning all projects, report:
-   ```
-   Unharvested knowledge:
-   - [Project A]: X entries
-   - [Project B]: Y entries
-   Run /harvest for any of these? (Yes [which ones] / No / After review)
-   ```
-5. **STOP. Wait for user response.** Do not proceed to 3D until the user answers.
-6. If user says yes: note which project(s) to harvest AFTER review-week completes
-7. If user says no or after: proceed
-
-**Project Status:**
-- Flag zombie projects: no activity in 14+ days → "Archive to `04_Archives/`?"
-- Flag stale projects: >30 days without activity → "Cold Project" warning
-- Flag orphan projects: no plan.md or project_state.md → "Missing state files"
-
-If any project flags raised: **STOP. Wait for user decision on each flag before proceeding.**
-
-### 3D. Strategy Pivot Check
-Compare week's inputs (from all Dailies) vs Strategy.md targets.
-
-**The Sample Size Protocol (MANDATORY):**
-Before recommending any pivot, check sample size.
-If sample < target (e.g., only 15/30 DMs sent):
-- "INSUFFICIENT DATA. You are in the 'Latency' period. Do not change variables. Keep going until N=[target]. Trust the math, ignore the feelings."
-
-**Verdict:**
-- **Green:** Metrics hit. Strategy validated. Keep going.
-- **Yellow:** Metrics close but inconsistent. Monitor another week.
-- **Red:** Metrics missed WITH sufficient sample. Root cause analysis needed.
-
-If Strategy Failed (Red + sufficient sample):
-- "Run `/pivot-strategy` to formally document the change."
-
-**Confidence labeling:**
-- "Correlation detected (Confidence: 3)" vs "Causation confirmed (Confidence: 5)"
-- Avoid blame-shifting. Find root causes, don't pin failures on the easiest variable.
-- Do NOT present single-day observations as weekly trends.
-- Require 3+ data points before claiming any causal relationship.
-- If data is insufficient: say "Insufficient data to determine root cause" rather than guessing.
-[Mover_Global_Rules.md: S2, S3]
-
-### 3E. Environmental & Input Scan (Identity Alignment)
-"Rate your environment this week (-10 to +10). Did your feed/friends/room fuel or drain you?"
-If negative: "Identify one Friction Point to remove and one Accelerant to add."
-
-**Input Audit (Weekly):**
-- **Content consumed:** What did the user read, watch, listen to this week? Did it reinforce the evolved identity or feed the Anti-Identity?
-- **Social circle:** Energy-givers vs energy-drainers encountered this week?
-- **Habitual inputs:** Any autopilot behaviors detected across multiple days? (Social media before Deep Work, defaulting to easy tasks, etc.)
-- Score: "Your inputs this week were approximately [X]% aligned with the identity you're building."
-- If < 50%: "Your environment is actively working against you. One structural change: [specific suggestion]."
-
-### 3F. Auto_Learnings Review (Score Decay + Domain TTLs)
-
-Review ALL entries in Auto_Learnings.md. Apply three decay mechanisms:
-
-**1. Evidence Decay (weekly):**
-- Entries with no new supporting evidence in 30+ days: reduce confidence by 0.5.
-- Entries at confidence ≤ 0: move to `## ARCHIVED (Decayed)` section at bottom of file.
-- Entries confirmed by new evidence this week: promote confidence by 0.5 (max 5).
-
-**2. Domain TTLs (time-to-live by category):**
-Entries have different shelf lives based on their domain:
-
-| Domain | TTL | Example |
-|--------|-----|---------|
-| Work habits | 90 days | "Avoids cold outreach", "Scope creeps on UI" |
-| Health/Vitality | 180 days | "Energy crashes at 2pm", "Skips gym on Mondays" |
-| Preferences | 365 days | "Prefers voice-to-text", "Uses dark mode" |
-| Identity/Values | No expiry | "Driven by proving doubters wrong" |
-| AI Corrections | 90 days | "AI uses && on Windows", "AI overwrites Engine files" |
-
-If an entry exceeds its TTL without reconfirmation: reduce confidence by 1 and flag for review.
-If expired AND confidence ≤ 1: archive automatically.
-
-**3. Pattern Lifecycle:**
-- Patterns absent 10+ days: reduce confidence by 1.
-- Patterns absent 21+ days: mark RESOLVED, archive.
-- Patterns confirmed 3+ times: promote confidence.
-- Chronic patterns (active 90+ days, confidence 4+): escalate — "This is structural, not situational."
-
-**4. CORRECTION Lifecycle:**
-- ACTIVE corrections absent 90+ days: reduce confidence by 1, set Status: MONITORING. "Fix appears to be working."
-- ACTIVE Conf 5 absent 90+ days: reduce to 4, flag: "Structural fix may have taken hold. Reducing severity."
-- DEFERRED corrections >30 days: surface explicitly — "Correction '[name]' deferred on [date]. Still relevant? Resolve or promote to ACTIVE."
-- RESOLVED corrections: no decay. Kept as permanent record.
-
-**Output:** Highlight top 3 active patterns for user awareness. Flag any entries that decayed or expired this week.
-
-### 3G. Wins Extraction
-From all Daily Notes and Session Logs, explicitly list wins:
-- Tasks completed that were carried 2+ days (zombie kills)
-- Metrics that hit or exceeded targets
-- Identity-aligned behavior (e.g., pipeline done before Mover OS)
-- System improvements shipped
-- Any moment of genuine momentum or breakthrough
-
-Output: "**Wins this week:** [List]" — this feeds directly into Step 7A's Weekly Review file.
-
-### 3H. System Improvement (Self-Evolving System)
-
-THIS is where the system evolves. Not just flagging issues — actually fixing them.
-
-**Source Scan (gather all issues):**
-1. Scan this week's Session Logs for "System Issue" flags
-2. Scan Auto_Learnings.md for patterns with confidence 3+ that suggest workflow/rule changes
-3. Scan this week's user corrections (from Self-Improvement sections in Session Logs)
-4. Group all issues by type: AI behavior | workflow gaps | rule additions | template fixes
-
-**Before proposing any fix, pass the Leverage Check:**
-1. Does the model already handle this without a rule? If yes: skip it.
-2. Is this enforcement or guidance? Enforcement → hook. Guidance → rule or workflow instruction.
-3. Does this apply globally or only in one workflow? One workflow → put it there, not in Global Rules.
-4. Does a similar rule already exist? If yes: strengthen the existing one, don't add a duplicate.
-
-**For EACH issue that passes the Leverage Check, present a concrete fix:**
-```
-ISSUE: [Description from source]
-SOURCE: [Session Log date / Auto_Learnings entry]
-FIX TYPE: [Workflow edit / Rule addition / Hook change / Template update]
-LEVERAGE CHECK: [Which level this targets and why]
-PROPOSED CHANGE:
-- File: [path]
-- OLD: "[current text]"
-+ NEW: "[proposed text]"
-IMPACT: [What this prevents/improves]
-```
-
-**STOP. Present all proposed fixes. Wait for user to approve/reject each one.**
-
-**For each approved fix:**
-1. Implement the change immediately (edit the file)
-2. Log in Weekly Review file: "System improvement: [fix] [date]"
-3. If it's a rule change: update Global Rules version number
-4. Append to Active_Context.md `## My Customizations`: `| [Date] | [File] | [What changed] | [Why] |`
-
-**For rejected fixes:** Note the rejection reason. If the same issue appears 3+ weeks, escalate: "This issue keeps recurring. Revisit?"
-
-This is how the system gets better every week. Issues identified → fixes proposed → fixes implemented → system evolves.
-
----
-
-## 4. SELF-VERIFICATION (CISC)
-
-Before outputting, self-check:
-1. All strategy verdicts cite weekly data with [file:line]? (90%+)
-2. Sample size validated before any pivot recommendation?
-3. Identity audit references specific daily evidence?
-4. Auto_Learnings decay rules applied?
-5. No single-day false causation presented as weekly trend?
-
-If any section < 90% confident: flag it.
-
----
-
-## 5. USER REPORT
-
-Present findings to the user BEFORE writing to any Engine files.
-
-"Here's what the week shows:"
-1. **Identity Score:** [X/10] — [one-line verdict citing specific days]
-2. **Strategy Verdict:** [Green/Yellow/Red] — [metric performance vs targets]
-3. **Wins:** [List from 3G]
-4. **Patterns:** [Top 3 from Auto_Learnings review]
-5. **System Issues:** [Count] issues found, [Count] fixes proposed
-6. **Inbox:** [Status]
-7. **Projects:** [Health summary]
-
-If MONTHLY_MODE: include monthly trends section.
-
-"Review this before I update Engine files. Any corrections?"
-
-WAIT FOR USER RESPONSE before proceeding to Engine updates.
-
----
-
-## 6. ENGINE FILE UPDATES (HIGH-RISK)
-
-This is where insights from the audit become permanent. Do not skip.
-
-**Diff format for all proposed edits:**
-```
-- OLD: "[exact current text]"
-+ NEW: "[proposed replacement]"
-Reason: [one line — what weekly data supports this change]
-```
-
-### 6A. Strategy.md Update
-
-Review Strategy.md every week regardless of verdict — even Green weeks can surface refinements.
-1. For **Green** verdict: check if any metric thresholds are too easy (tighten if needed)
-2. For **Yellow** verdict: propose constraint or variable adjustments
-3. For **Red** verdict + sufficient sample: propose pivot criteria update or route to `/pivot-strategy`
-4. If full pivot needed: run `/pivot-strategy` first, then return here
-
-"Proposed Strategy.md changes:
-[DIFF FORMAT ABOVE]
-Confirm? [Yes / Skip / Run /pivot-strategy instead]"
-
-WAIT FOR USER CONFIRMATION before writing. [Mover_Global_Rules.md: S3 HIGH-RISK]
-
-### 6B. Identity_Prime.md Update
-
-Review Identity_Prime.md every week — both for drift corrections AND proactive additions.
-1. **Drift corrections** (if 3A found slippage): sharpen Anti-Identity with specific failure mode observed
-2. **Proactive updates** (always check): any new rituals started? Energy window shifted? New non-negotiable emerged?
-3. Do NOT rewrite core values — append observations only
-4. Vitality/Faith: check if sleep patterns, spiritual practices, or energy data from the week suggest any update to these sections of Identity_Prime.md
-
-"Proposed Identity_Prime.md changes:
-[DIFF FORMAT ABOVE]
-Confirm? [Yes / Skip]"
-
-WAIT FOR USER CONFIRMATION. Update `**Last Updated:**` date if approved.
-
-### 6C. Active_Context.md Refresh
-
-Always runs — Active_Context is always stale after a week.
-1. What changed in strategic state this week?
-2. Any blockers resolved? Add new ones?
-3. Update "Current Phase" if shifted
-4. Section 2 (Graveyard): if any ideas were killed this week (from Inbox triage, Someday/Maybe pruning, or user decisions), add them here with reason and date
-5. Section 4 (Technical Context): if active project changed, update project name, path, and tech stack
-6. Section 6 (Active Sessions): verify buffer is current — trim to last 5 entries
-7. Write workflow state: `weekly_review_last_run: [YYYY-MM-DD]`
-8. If MONTHLY_MODE was true: also write `monthly_review_last_run: [YYYY-MM-DD]` to Workflow State
-9. **Identity Snapshot (S5):** Refresh from Identity_Prime.md — Archetype, Core Drive, The Code (top 3), Anti-Identity (full triggers), Current Ritual. This is the cache that light workflows read instead of the full Identity file.
-10. **Active Patterns (S5.5):** Extract all patterns from Auto_Learnings.md at Confidence 3+. Format: `[PATTERN/CORRECTION] [Description] (Conf [X], [SEVERITY]) — Suggested: /[workflow]`. Light workflows read this instead of the full Auto_Learnings file.
-
-"Proposed Active_Context.md changes:
-[DIFF FORMAT — only the lines that change, not the full file]
-Confirm? [Yes / Edit / Skip]"
-
-WAIT FOR USER CONFIRMATION.
-
-### 6D. Git Commit
-
-After all Engine file edits are confirmed:
-```
-git add 02_Areas/Engine/
-git commit -m "chore(engine): weekly review updates — [YYYY-MM-DD]"
-```
-Do NOT commit `Weekly Reviews/` or `Dailies/` — these are gitignored. [Mover_Global_Rules.md: S9]
-
----
-
-## 7. THE RESET & PLAN
-
-### 7A. Weekly Review File
-Create `{VAULT_ROOT}/02_Areas/Engine/Weekly Reviews/Week - [YYYY-WW].md` with:
-- Identity Score: [X]/10
-- Strategy Status: [Valid / Monitor / Pivot Needed]
-- Key Patterns: [Top 3 from Auto_Learnings]
-- Wins: [List]
-- Failures: [List with root cause]
-- Engine Files Updated: [List what changed]
-- System Improvements Made: [List]
-
-### 7B. Next Week's Objectives
-Output the **3 Big Objectives** for next week based on Strategy and this review.
-
-### 7C. Monthly Mode (Every 4th Week)
-If this is the 4th weekly review in a row, trigger deeper reflection.
-
-**Batch Reading Protocol (Monthly):**
-1. Read ALL Daily Notes from this month — split into batches of 5-7 files
-2. After each batch, extract: completion rates, energy trends, pattern frequencies, strategy adherence
-3. Synthesize across ALL batches before drawing conclusions
-4. Cross-reference with ALL Engine files: Identity_Prime, Strategy, Active_Context, Goals, Dossier, Metrics_Log, Auto_Learnings
-5. This is the deepest analysis the system does. Do NOT be shallow. Read everything.
-
-**Baseline Detection (Month 1 Only):**
-If Metrics_Log.md has NO `## Baseline` section AND this is the first monthly review with 20+ daily entries:
-1. Calculate averages from the month's data: completion rate, deep work hours, energy, single test hit rate, zombie task count
-2. Write a `## Baseline` section to Metrics_Log.md with these averages
-3. "This is your Month 1 baseline. From next month, the system tracks improvement against these numbers."
-4. Ask user: "Do these numbers feel accurate? Any anomalies this month that skew them?"
-
-If `## Baseline` already exists: compare this month's averages to baseline. Output delta: "Completion rate: 72% (baseline: 58%, +14%). Deep work: 4.2h (baseline: 3.1h, +1.1h)."
-
-**Monthly Reflection:**
-- Identity Audit: Which traits achieved? Which still in progress? Read Identity_Prime.md and compare against the month's behavior.
-- Dossier Review: Any new assets, skills, or connections to add? Read Mover_Dossier.md.
-- Goals Review: Are goals still aligned with identity? Are 90-day goals on track? Read Goals.md.
-- Strategy evolution: How has Strategy.md changed over the month? Were pivots justified by data?
-- Pattern lifecycle: Which Auto_Learnings patterns were resolved? Which are chronic?
-- Subconscious patterns: Any recurring behavioral patterns the user should be aware of? (Avoidance, energy, scope creep)
-- Baseline comparison: If baseline exists, how does this month compare? Show trends.
-
-**Output:** Create `{VAULT_ROOT}/02_Areas/Engine/Monthly Reviews/Month - [YYYY-MM].md`. Separate from the weekly review file. Include month-over-month trends and baseline comparison.
-
-### 7D. Quarterly Mode (Every Quarter End)
-
-If QUARTERLY_MODE=true, this is the deepest analysis cycle. Everything in Monthly Mode runs PLUS:
-
-**90-Day Goal Review:**
-1. Read Goals.md. For each 90-day goal:
-   - What was the target? What was achieved? Show the gap as a number.
-   - Was the goal realistic given the data? Should it be adjusted, kept, or replaced?
-   - If a goal was missed: root cause analysis. Was it strategy failure, execution failure, or goal-setting failure?
-2. Set new 90-day goals (or carry forward adjusted ones). Require user confirmation.
-
-**Strategy Evolution Audit:**
-1. Read Strategy.md CHANGELOG. Map every pivot this quarter.
-2. Were pivots data-driven or emotional? Count each type.
-3. Is the current hypothesis still valid given 3 months of data?
-4. "Your strategy changed [X] times this quarter. [Y] were data-driven, [Z] were hypothesis-only. This tells us: [verdict]."
-
-**Identity Trajectory:**
-1. Compare Identity_Prime.md from start of quarter (via git history if available) to current.
-2. What evolved? What regressed? What stayed constant?
-3. "Three months ago, you described yourself as [X]. The data shows you've become [Y]. Gap: [Z]."
-
-**Dossier Review:**
-1. Read Mover_Dossier.md. What assets were gained this quarter? What depreciated?
-2. New skills, connections, capital, content assets?
-3. "Your leverage position changed from [start] to [end]. Net: [gain/loss]."
-
-**Auto_Learnings Lifecycle:**
-1. Which patterns were resolved this quarter?
-2. Which became chronic (active for 3+ months)?
-3. Chronic patterns get escalated: "Pattern [X] has been active for [Y] months. This is now structural, not situational. Needs a system-level fix, not willpower."
-
-**Rules Health Check (Global Rules audit):**
-1. Read `Mover_Global_Rules.md` section by section.
-2. For each rule, ask: "Did this fire this quarter? Did it prevent a real mistake, or is it dead weight?"
-3. Flag rules that: (a) the model now handles natively without instruction, (b) duplicate another rule, (c) could be a hook instead.
-4. Propose removals or consolidations. Every line in Global Rules costs instruction budget.
-5. Present findings: "Global Rules: [X] sections, [Y] lines. Proposed cuts: [list]. Net savings: [Z] lines."
-6. Wait for user approval before any changes.
-
-**Quarterly Output:**
-Create `{VAULT_ROOT}/02_Areas/Engine/Quarterly Reviews/Quarter - [YYYY-QN].md` with:
-- 90-day goal scorecard
-- Strategy evolution timeline
-- Identity trajectory
-- Leverage audit
-- Chronic pattern list
-- Rules health findings
-- Next quarter's 90-day goals (proposed)
-
-"Review this quarterly report. Then run `/plan-tomorrow` for the first day of the new quarter."
-
----
-
-## 8. SELF-IMPROVEMENT
-
-If user corrected AI behavior during this workflow:
-
-1. Acknowledge: "I made a mistake."
-2. Identify the violated rule AND the source file.
-3. Propose fix using S11 diff format:
-   ```
-   CORRECTION FIX:
-   - File: [path]
-   - OLD: "[current text]"
-   + NEW: "[proposed text]"
-   SAFETY CHECK: [Which Safe Modification gate this passes]
-   ```
-4. **STOP. Wait for user approval.** (Same gate as S11 Trigger A.)
-5. If approved: implement + log structured CORRECTION to Auto_Learnings.md (see S11 format).
-6. If deferred: log with `Status: DEFERRED`.
-7. Before proposing rule additions, pass the Rule Addition Gate (Global Rules S11).
-8. If same correction seen 3x+, escalate confidence to 5 and flag as structural.
-
----
-
-## 9. PROACTIVE SUGGESTIONS
-
-Based on weekly observations:
-1. [Suggestion with confidence score]
-2. [Suggestion with confidence score]
-3. [Suggestion with confidence score]
-
-"Want me to address any of these?"
-
----
-
-## 10. ANSWER GATING
-
-Maintain an internal checklist. Before every response, check which gates are resolved:
-- [ ] Strategy verdict delivered and acknowledged
-- [ ] Pivot decision made or deferred
-- [ ] Engine files updated or explicitly skipped (Active_Context is NOT skippable)
-- [ ] Inbox processed or acknowledged as skipped
-- [ ] System improvements discussed
-- [ ] Wins list generated (Step 3G)
-
-If any gate is open: "You haven't confirmed [X]. Wait or proceed?"
-
----
-
-## HANDOFF
-
-Sources used: [List ALL files read]
-Confidence range: [X-Y]
-
-**Next:** "Weekly Review complete. Ready to run `/plan-tomorrow` for Monday?"
-If Strategy Pivot needed: "First run `/pivot-strategy`, then `/plan-tomorrow`."
-
-Key rules still active:
-- All claims must be grounded
-- Strategy changes require user confirmation
-- Sample size must be met before recommending pivots
-- Append-only: never delete strategy history
+---
+description: Weekly review. Clean the system, harvest wisdom, audit identity, validate strategy, implement system improvements.
+---
+
+# Weekly Review (V4)
+
+**Trigger:** User runs `/review-week` (user-configured day, default Sunday).
+**Role:** Chief of Staff & Performance Psychologist.
+**Task:** Clean system, harvest wisdom, audit identity, validate strategy, reset for next week.
+**Risk Tier:** HIGH (Strategy changes require confirmation)
+
+This is the system improvement point. Issues detected during the week get fixed here.
+
+---
+
+## 0. PRE-FLIGHT
+
+Global Rules are pre-loaded as CLAUDE.md / GEMINI.md. Output `[Rules Loaded - DATE]`.
+
+VAULT ROOT (resolve ONCE, use for all file paths):
+```bash
+git rev-parse --show-toplevel
+```
+Save as VAULT_ROOT. All paths below are relative to VAULT_ROOT.
+
+**Review file paths:**
+- Weekly: `{VAULT_ROOT}/02_Areas/Engine/Weekly Reviews/Week - [YYYY-WW].md`
+- Monthly: `{VAULT_ROOT}/02_Areas/Engine/Monthly Reviews/Month - [YYYY-MM].md`
+- Quarterly: `{VAULT_ROOT}/02_Areas/Engine/Quarterly Reviews/Quarter - [YYYY-QN].md`
+
+Create the folder if it doesn't exist before writing. Never write outside `02_Areas/Engine/`.
+
+LOAD (Memory Hierarchy):
+1. Auto_Learnings.md (P0) - FULL file
+2. Active_Context.md (P1)
+3. Strategy.md (P2)
+4. Identity_Prime.md (P3)
+5. Mover_Dossier.md (P3)
+6. Goals.md (if exists)
+7. Metrics_Log.md (if exists) — 30-day rolling data for strategy verdict
+8. Daily Notes from this week — read on demand in Step 2A, not all upfront (P5)
+
+REVIEW DAY: Read `review_week_day` from Active_Context.md Workflow State (default: Sunday). All "review day" references in this workflow use this user-configured day.
+
+MONTHLY MODE CHECK (date-based, run before proceeding):
+```bash
+# Is today the last occurrence of review_week_day this month?
+DAYS_LEFT=$(python3 -c "
+from datetime import date, timedelta
+today = date.today()
+next_week = today + timedelta(days=7)
+print('LAST' if next_week.month != today.month else 'NOT_LAST')
+" 2>/dev/null || node -e "
+const t=new Date(), n=new Date(t); n.setDate(t.getDate()+7);
+console.log(n.getMonth()!==t.getMonth()?'LAST':'NOT_LAST');
+")
+```
+If output is `LAST`: set MONTHLY_MODE=true. This is the user's monthly review day.
+
+QUARTERLY MODE CHECK (date-based):
+If MONTHLY_MODE=true, also check if current month is March, June, September, or December.
+If yes: set QUARTERLY_MODE=true.
+
+CONSTRAINTS:
+- All verdicts grounded in weekly data, not single-day observations
+- Strategy changes require user confirmation (HIGH-RISK)
+- Append-only: never delete strategy history
+- Sample size validation before recommending pivots
+- Label causal claims with confidence level
+- Windows: no && in commands
+
+---
+
+## 1. PREREQUISITE CHECK
+
+- At least 3 Daily Notes exist for this week (minimum data for review)
+- Strategy.md exists
+- Identity_Prime.md exists
+- Today's `/analyse-day` should be run first (suggest if not)
+
+If < 3 Daily Notes: CAUTION. "Only [X] days logged. Review will be limited. Proceed anyway?"
+
+---
+
+## 2. THE SYSTEM AUDIT
+
+### 2A. Read All Sources
+Using files already loaded in Pre-Flight, extract the following. Only re-read if content not yet in context:
+1. Strategy.md — Hypothesis, Metrics, Sample Sizes, Pivot Criteria
+2. Identity_Prime.md — Values, Anti-Identity, Non-negotiables, Rituals
+3. This week's Daily Notes — read one at a time, extracting: Tasks completed, Battery scores, Session Log events, Key Insights
+4. Auto_Learnings.md — full pattern database (already loaded)
+5. Goals.md — current targets (if loaded)
+6. Metrics_Log.md — last 30 days of data (if loaded)
+7. SCAN `00_Inbox/` — Quick Capture, Integrations folder
+8. SCAN `01_Projects/` — list active projects, last commit dates
+
+After each Daily Note read, output a one-line extract: `[Date]: Tasks X/Y | Energy: Z | Notable: ...`
+This keeps you from losing early-week data by the time you reach later entries.
+
+### 2B. Batch Reading Protocol
+If > 7 files to read:
+1. Split into batches of 4-5 files
+2. Read each batch, extract patterns
+3. Synthesize across batches
+4. Cross-reference with Engine files
+
+---
+
+## 3. THE PROCESSING PROTOCOL
+
+### 3A. Identity Audit (The Mirror + Evolved Self Tracking)
+Compare user's behavior across ALL dailies vs Identity_Prime.md.
+- "Did we embody the identity this week, or did we slip into the Anti-Identity?"
+- If slipped: identify trigger (Lack of sleep? Fear? Overwhelm?) and prescribe fix.
+- Cite specific days: [Daily - YYYY-MM-DD] for evidence.
+- Do NOT soften the verdict. If the week was bad, say it was bad. "You planned to be [identity] this week. The data shows you were [actual behavior] on [X] out of [Y] days. That's not a rough week, that's a pattern."
+- If the same slippage appeared last week: "This is the second consecutive week of [pattern]. Once is a bad week. Twice is a habit forming. What structurally changes next week?"
+
+**Subconscious Pattern Tracking:**
+- Which evolved behaviors are becoming automatic? (Happened without conscious effort this week)
+- Which still require conscious effort? (Had to force yourself to do them)
+- Any new Anti-Identity triggers surfaced this week that aren't in Identity_Prime.md yet?
+- Track the trajectory: "Week over week, [behavior X] is becoming more natural / still requires effort."
+
+### 3B. Inbox Zero Check
+Scan ALL files in `{VAULT_ROOT}/00_Inbox/`. For each item, classify and route:
+
+| Content Type | Destination |
+|---|---|
+| Task / action item | Active project's plan.md or Active_Context Backlog |
+| Article / video / book note | `03_Library/Inputs/[Articles\|Videos\|Books]/` using template |
+| Reusable knowledge (pattern, principle, SOP) | `03_Library/` (Cheatsheets/Principles/SOPs) |
+| Person / org / place mentioned | `03_Library/Entities/[type]/` using template |
+| Decision documented | `03_Library/Entities/Decisions/` using template |
+| Project idea (not started) | `02_Areas/Engine/Someday_Maybe.md` |
+| Noise / outdated / irrelevant | Delete or move to `04_Archives/` |
+
+Process `00_Inbox/Quick_Capture.md` (if exists): parse each entry and route individually.
+Process `00_Inbox/Integrations/` folder if it exists.
+
+**Goal: Inbox must be empty by the end of review.** Every item either routed or trashed.
+
+If > 10 items: batch-present them to the user for quick triage rather than asking one by one.
+
+### 3C. Project Health & Harvest Scan
+For each active project in `01_Projects/`:
+
+**Harvest Scan (DO NOT harvest inline — just count):**
+1. Read `project_state.md` (if exists)
+2. Count Solutions Ledger entries NOT in `## ARCHIVED KNOWLEDGE (Harvested)`
+3. If unharvested entries > 0: record project name + count
+4. After scanning all projects, report:
+   ```
+   Unharvested knowledge:
+   - [Project A]: X entries
+   - [Project B]: Y entries
+   Run /harvest for any of these? (Yes [which ones] / No / After review)
+   ```
+5. **STOP. Wait for user response.** Do not proceed to 3D until the user answers.
+6. If user says yes: note which project(s) to harvest AFTER review-week completes
+7. If user says no or after: proceed
+
+**Project Status:**
+- Flag zombie projects: no activity in 14+ days → "Archive to `04_Archives/`?"
+- Flag stale projects: >30 days without activity → "Cold Project" warning
+- Flag orphan projects: no plan.md or project_state.md → "Missing state files"
+
+If any project flags raised: **STOP. Wait for user decision on each flag before proceeding.**
+
+### 3D. Strategy Pivot Check
+Compare week's inputs (from all Dailies) vs Strategy.md targets.
+
+**The Sample Size Protocol (MANDATORY):**
+Before recommending any pivot, check sample size.
+If sample < target (e.g., only 15/30 DMs sent):
+- "INSUFFICIENT DATA. You are in the 'Latency' period. Do not change variables. Keep going until N=[target]. Trust the math, ignore the feelings."
+
+**Verdict:**
+- **Green:** Metrics hit. Strategy validated. Keep going.
+- **Yellow:** Metrics close but inconsistent. Monitor another week.
+- **Red:** Metrics missed WITH sufficient sample. Root cause analysis needed.
+
+If Strategy Failed (Red + sufficient sample):
+- "Run `/pivot-strategy` to formally document the change."
+
+**Confidence labeling:**
+- "Correlation detected (Confidence: 3)" vs "Causation confirmed (Confidence: 5)"
+- Avoid blame-shifting. Find root causes, don't pin failures on the easiest variable.
+- Do NOT present single-day observations as weekly trends.
+- Require 3+ data points before claiming any causal relationship.
+- If data is insufficient: say "Insufficient data to determine root cause" rather than guessing.
+[Mover_Global_Rules.md: S2, S3]
+
+### 3E. Environmental & Input Scan (Identity Alignment)
+"Rate your environment this week (-10 to +10). Did your feed/friends/room fuel or drain you?"
+If negative: "Identify one Friction Point to remove and one Accelerant to add."
+
+**Input Audit (Weekly):**
+- **Content consumed:** What did the user read, watch, listen to this week? Did it reinforce the evolved identity or feed the Anti-Identity?
+- **Social circle:** Energy-givers vs energy-drainers encountered this week?
+- **Habitual inputs:** Any autopilot behaviors detected across multiple days? (Social media before Deep Work, defaulting to easy tasks, etc.)
+- Score: "Your inputs this week were approximately [X]% aligned with the identity you're building."
+- If < 50%: "Your environment is actively working against you. One structural change: [specific suggestion]."
+
+### 3F. Auto_Learnings Review (Score Decay + Domain TTLs)
+
+Review ALL entries in Auto_Learnings.md. Apply three decay mechanisms:
+
+**1. Evidence Decay (weekly):**
+- Entries with no new supporting evidence in 30+ days: reduce confidence by 0.5.
+- Entries at confidence ≤ 0: move to `## ARCHIVED (Decayed)` section at bottom of file.
+- Entries confirmed by new evidence this week: promote confidence by 0.5 (max 5).
+
+**2. Domain TTLs (time-to-live by category):**
+Entries have different shelf lives based on their domain:
+
+| Domain | TTL | Example |
+|--------|-----|---------|
+| Work habits | 90 days | "Avoids cold outreach", "Scope creeps on UI" |
+| Health/Vitality | 180 days | "Energy crashes at 2pm", "Skips gym on Mondays" |
+| Preferences | 365 days | "Prefers voice-to-text", "Uses dark mode" |
+| Identity/Values | No expiry | "Driven by proving doubters wrong" |
+| AI Corrections | 90 days | "AI uses && on Windows", "AI overwrites Engine files" |
+
+If an entry exceeds its TTL without reconfirmation: reduce confidence by 1 and flag for review.
+If expired AND confidence ≤ 1: archive automatically.
+
+**3. Pattern Lifecycle:**
+- Patterns absent 10+ days: reduce confidence by 1.
+- Patterns absent 21+ days: mark RESOLVED, archive.
+- Patterns confirmed 3+ times: promote confidence.
+- Chronic patterns (active 90+ days, confidence 4+): escalate — "This is structural, not situational."
+
+**4. CORRECTION Lifecycle:**
+- ACTIVE corrections absent 90+ days: reduce confidence by 1, set Status: MONITORING. "Fix appears to be working."
+- ACTIVE Conf 5 absent 90+ days: reduce to 4, flag: "Structural fix may have taken hold. Reducing severity."
+- DEFERRED corrections >30 days: surface explicitly — "Correction '[name]' deferred on [date]. Still relevant? Resolve or promote to ACTIVE."
+- RESOLVED corrections: no decay. Kept as permanent record.
+
+**Output:** Highlight top 3 active patterns for user awareness. Flag any entries that decayed or expired this week.
+
+### 3G. Wins Extraction
+From all Daily Notes and Session Logs, explicitly list wins:
+- Tasks completed that were carried 2+ days (zombie kills)
+- Metrics that hit or exceeded targets
+- Identity-aligned behavior (e.g., pipeline done before Mover OS)
+- System improvements shipped
+- Any moment of genuine momentum or breakthrough
+
+Output: "**Wins this week:** [List]" — this feeds directly into Step 7A's Weekly Review file.
+
+### 3H. System Improvement (Self-Evolving System)
+
+THIS is where the system evolves. Not just flagging issues — actually fixing them.
+
+**Source Scan (gather all issues):**
+1. Scan this week's Session Logs for "System Issue" flags
+2. Scan Auto_Learnings.md for patterns with confidence 3+ that suggest workflow/rule changes
+3. Scan this week's user corrections (from Self-Improvement sections in Session Logs)
+4. Group all issues by type: AI behavior | workflow gaps | rule additions | template fixes
+
+**Before proposing any fix, pass the Leverage Check:**
+1. Does the model already handle this without a rule? If yes: skip it.
+2. Is this enforcement or guidance? Enforcement → hook. Guidance → rule or workflow instruction.
+3. Does this apply globally or only in one workflow? One workflow → put it there, not in Global Rules.
+4. Does a similar rule already exist? If yes: strengthen the existing one, don't add a duplicate.
+
+**For EACH issue that passes the Leverage Check, present a concrete fix:**
+```
+ISSUE: [Description from source]
+SOURCE: [Session Log date / Auto_Learnings entry]
+FIX TYPE: [Workflow edit / Rule addition / Hook change / Template update]
+LEVERAGE CHECK: [Which level this targets and why]
+PROPOSED CHANGE:
+- File: [path]
+- OLD: "[current text]"
++ NEW: "[proposed text]"
+IMPACT: [What this prevents/improves]
+```
+
+**STOP. Present all proposed fixes. Wait for user to approve/reject each one.**
+
+**For each approved fix:**
+1. Implement the change immediately (edit the file)
+2. Log in Weekly Review file: "System improvement: [fix] [date]"
+3. If it's a rule change: update Global Rules version number
+4. Append to Active_Context.md `## My Customizations`: `| [Date] | [File] | [What changed] | [Why] |`
+
+**For rejected fixes:** Note the rejection reason. If the same issue appears 3+ weeks, escalate: "This issue keeps recurring. Revisit?"
+
+This is how the system gets better every week. Issues identified → fixes proposed → fixes implemented → system evolves.
+
+---
+
+## 4. SELF-VERIFICATION (CISC)
+
+Before outputting, self-check:
+1. All strategy verdicts cite weekly data with [file:line]? (90%+)
+2. Sample size validated before any pivot recommendation?
+3. Identity audit references specific daily evidence?
+4. Auto_Learnings decay rules applied?
+5. No single-day false causation presented as weekly trend?
+
+If any section < 90% confident: flag it.
+
+---
+
+## 5. USER REPORT
+
+Present findings to the user BEFORE writing to any Engine files.
+
+"Here's what the week shows:"
+1. **Identity Score:** [X/10] — [one-line verdict citing specific days]
+2. **Strategy Verdict:** [Green/Yellow/Red] — [metric performance vs targets]
+3. **Wins:** [List from 3G]
+4. **Patterns:** [Top 3 from Auto_Learnings review]
+5. **System Issues:** [Count] issues found, [Count] fixes proposed
+6. **Inbox:** [Status]
+7. **Projects:** [Health summary]
+
+If MONTHLY_MODE: include monthly trends section.
+
+"Review this before I update Engine files. Any corrections?"
+
+WAIT FOR USER RESPONSE before proceeding to Engine updates.
+
+---
+
+## 6. ENGINE FILE UPDATES (HIGH-RISK)
+
+This is where insights from the audit become permanent. Do not skip.
+
+**Diff format for all proposed edits:**
+```
+- OLD: "[exact current text]"
++ NEW: "[proposed replacement]"
+Reason: [one line — what weekly data supports this change]
+```
+
+### 6A. Strategy.md Update
+
+Review Strategy.md every week regardless of verdict — even Green weeks can surface refinements.
+1. For **Green** verdict: check if any metric thresholds are too easy (tighten if needed)
+2. For **Yellow** verdict: propose constraint or variable adjustments
+3. For **Red** verdict + sufficient sample: propose pivot criteria update or route to `/pivot-strategy`
+4. If full pivot needed: run `/pivot-strategy` first, then return here
+
+"Proposed Strategy.md changes:
+[DIFF FORMAT ABOVE]
+Confirm? [Yes / Skip / Run /pivot-strategy instead]"
+
+WAIT FOR USER CONFIRMATION before writing. [Mover_Global_Rules.md: S3 HIGH-RISK]
+
+### 6B. Identity_Prime.md Update
+
+Review Identity_Prime.md every week — both for drift corrections AND proactive additions.
+1. **Drift corrections** (if 3A found slippage): sharpen Anti-Identity with specific failure mode observed
+2. **Proactive updates** (always check): any new rituals started? Energy window shifted? New non-negotiable emerged?
+3. Do NOT rewrite core values — append observations only
+4. Vitality/Faith: check if sleep patterns, spiritual practices, or energy data from the week suggest any update to these sections of Identity_Prime.md
+
+"Proposed Identity_Prime.md changes:
+[DIFF FORMAT ABOVE]
+Confirm? [Yes / Skip]"
+
+WAIT FOR USER CONFIRMATION. Update `**Last Updated:**` date if approved.
+
+### 6C. Active_Context.md Refresh
+
+Always runs — Active_Context is always stale after a week.
+1. What changed in strategic state this week?
+2. Any blockers resolved? Add new ones?
+3. Update "Current Phase" if shifted
+4. Section 2 (Graveyard): if any ideas were killed this week (from Inbox triage, Someday/Maybe pruning, or user decisions), add them here with reason and date
+5. Section 4 (Technical Context): if active project changed, update project name, path, and tech stack
+6. Section 6 (Active Sessions): verify buffer is current — trim to last 5 entries
+7. Write workflow state: `weekly_review_last_run: [YYYY-MM-DD]`
+8. If MONTHLY_MODE was true: also write `monthly_review_last_run: [YYYY-MM-DD]` to Workflow State
+9. **Identity Snapshot (S5):** Refresh from Identity_Prime.md — Archetype, Core Drive, The Code (top 3), Anti-Identity (full triggers), Current Ritual. This is the cache that light workflows read instead of the full Identity file.
+10. **Active Patterns (S5.5):** Extract all patterns from Auto_Learnings.md at Confidence 3+. Format: `[PATTERN/CORRECTION] [Description] (Conf [X], [SEVERITY]) — Suggested: /[workflow]`. Light workflows read this instead of the full Auto_Learnings file.
+
+"Proposed Active_Context.md changes:
+[DIFF FORMAT — only the lines that change, not the full file]
+Confirm? [Yes / Edit / Skip]"
+
+WAIT FOR USER CONFIRMATION.
+
+### 6D. Git Commit
+
+After all Engine file edits are confirmed:
+```
+git add 02_Areas/Engine/
+git commit -m "chore(engine): weekly review updates — [YYYY-MM-DD]"
+```
+Do NOT commit `Weekly Reviews/` or `Dailies/` — these are gitignored. [Mover_Global_Rules.md: S9]
+
+---
+
+## 7. THE RESET & PLAN
+
+### 7A. Weekly Review File
+Create `{VAULT_ROOT}/02_Areas/Engine/Weekly Reviews/Week - [YYYY-WW].md` with:
+- Identity Score: [X]/10
+- Strategy Status: [Valid / Monitor / Pivot Needed]
+- Key Patterns: [Top 3 from Auto_Learnings]
+- Wins: [List]
+- Failures: [List with root cause]
+- Engine Files Updated: [List what changed]
+- System Improvements Made: [List]
+
+### 7B. Next Week's Objectives
+Output the **3 Big Objectives** for next week based on Strategy and this review.
+
+### 7C. Monthly Mode (Every 4th Week)
+If this is the 4th weekly review in a row, trigger deeper reflection.
+
+**Batch Reading Protocol (Monthly):**
+1. Read ALL Daily Notes from this month — split into batches of 5-7 files
+2. After each batch, extract: completion rates, energy trends, pattern frequencies, strategy adherence
+3. Synthesize across ALL batches before drawing conclusions
+4. Cross-reference with ALL Engine files: Identity_Prime, Strategy, Active_Context, Goals, Dossier, Metrics_Log, Auto_Learnings
+5. This is the deepest analysis the system does. Do NOT be shallow. Read everything.
+
+**Baseline Detection (Month 1 Only):**
+If Metrics_Log.md has NO `## Baseline` section AND this is the first monthly review with 20+ daily entries:
+1. Calculate averages from the month's data: completion rate, deep work hours, energy, single test hit rate, zombie task count
+2. Write a `## Baseline` section to Metrics_Log.md with these averages
+3. "This is your Month 1 baseline. From next month, the system tracks improvement against these numbers."
+4. Ask user: "Do these numbers feel accurate? Any anomalies this month that skew them?"
+
+If `## Baseline` already exists: compare this month's averages to baseline. Output delta: "Completion rate: 72% (baseline: 58%, +14%). Deep work: 4.2h (baseline: 3.1h, +1.1h)."
+
+**Monthly Reflection:**
+- Identity Audit: Which traits achieved? Which still in progress? Read Identity_Prime.md and compare against the month's behavior.
+- Dossier Review: Any new assets, skills, or connections to add? Read Mover_Dossier.md.
+- Goals Review: Are goals still aligned with identity? Are 90-day goals on track? Read Goals.md.
+- Strategy evolution: How has Strategy.md changed over the month? Were pivots justified by data?
+- Pattern lifecycle: Which Auto_Learnings patterns were resolved? Which are chronic?
+- Subconscious patterns: Any recurring behavioral patterns the user should be aware of? (Avoidance, energy, scope creep)
+- Baseline comparison: If baseline exists, how does this month compare? Show trends.
+
+**Output:** Create `{VAULT_ROOT}/02_Areas/Engine/Monthly Reviews/Month - [YYYY-MM].md`. Separate from the weekly review file. Include month-over-month trends and baseline comparison.
+
+### 7D. Quarterly Mode (Every Quarter End)
+
+If QUARTERLY_MODE=true, this is the deepest analysis cycle. Everything in Monthly Mode runs PLUS:
+
+**90-Day Goal Review:**
+1. Read Goals.md. For each 90-day goal:
+   - What was the target? What was achieved? Show the gap as a number.
+   - Was the goal realistic given the data? Should it be adjusted, kept, or replaced?
+   - If a goal was missed: root cause analysis. Was it strategy failure, execution failure, or goal-setting failure?
+2. Set new 90-day goals (or carry forward adjusted ones). Require user confirmation.
+
+**Strategy Evolution Audit:**
+1. Read Strategy.md CHANGELOG. Map every pivot this quarter.
+2. Were pivots data-driven or emotional? Count each type.
+3. Is the current hypothesis still valid given 3 months of data?
+4. "Your strategy changed [X] times this quarter. [Y] were data-driven, [Z] were hypothesis-only. This tells us: [verdict]."
+
+**Identity Trajectory:**
+1. Compare Identity_Prime.md from start of quarter (via git history if available) to current.
+2. What evolved? What regressed? What stayed constant?
+3. "Three months ago, you described yourself as [X]. The data shows you've become [Y]. Gap: [Z]."
+
+**Dossier Review:**
+1. Read Mover_Dossier.md. What assets were gained this quarter? What depreciated?
+2. New skills, connections, capital, content assets?
+3. "Your leverage position changed from [start] to [end]. Net: [gain/loss]."
+
+**Auto_Learnings Lifecycle:**
+1. Which patterns were resolved this quarter?
+2. Which became chronic (active for 3+ months)?
+3. Chronic patterns get escalated: "Pattern [X] has been active for [Y] months. This is now structural, not situational. Needs a system-level fix, not willpower."
+
+**Rules Health Check (Global Rules audit):**
+1. Read `Mover_Global_Rules.md` section by section.
+2. For each rule, ask: "Did this fire this quarter? Did it prevent a real mistake, or is it dead weight?"
+3. Flag rules that: (a) the model now handles natively without instruction, (b) duplicate another rule, (c) could be a hook instead.
+4. Propose removals or consolidations. Every line in Global Rules costs instruction budget.
+5. Present findings: "Global Rules: [X] sections, [Y] lines. Proposed cuts: [list]. Net savings: [Z] lines."
+6. Wait for user approval before any changes.
+
+**Quarterly Output:**
+Create `{VAULT_ROOT}/02_Areas/Engine/Quarterly Reviews/Quarter - [YYYY-QN].md` with:
+- 90-day goal scorecard
+- Strategy evolution timeline
+- Identity trajectory
+- Leverage audit
+- Chronic pattern list
+- Rules health findings
+- Next quarter's 90-day goals (proposed)
+
+"Review this quarterly report. Then run `/plan-tomorrow` for the first day of the new quarter."
+
+---
+
+## 8. SELF-IMPROVEMENT
+
+If user corrected AI behavior during this workflow:
+
+1. Acknowledge: "I made a mistake."
+2. Identify the violated rule AND the source file.
+3. Propose fix using S11 diff format:
+   ```
+   CORRECTION FIX:
+   - File: [path]
+   - OLD: "[current text]"
+   + NEW: "[proposed text]"
+   SAFETY CHECK: [Which Safe Modification gate this passes]
+   ```
+4. **STOP. Wait for user approval.** (Same gate as S11 Trigger A.)
+5. If approved: implement + log structured CORRECTION to Auto_Learnings.md (see S11 format).
+6. If deferred: log with `Status: DEFERRED`.
+7. Before proposing rule additions, pass the Rule Addition Gate (Global Rules S11).
+8. If same correction seen 3x+, escalate confidence to 5 and flag as structural.
+
+---
+
+## 9. PROACTIVE SUGGESTIONS
+
+Based on weekly observations:
+1. [Suggestion with confidence score]
+2. [Suggestion with confidence score]
+3. [Suggestion with confidence score]
+
+"Want me to address any of these?"
+
+---
+
+## 10. ANSWER GATING
+
+Maintain an internal checklist. Before every response, check which gates are resolved:
+- [ ] Strategy verdict delivered and acknowledged
+- [ ] Pivot decision made or deferred
+- [ ] Engine files updated or explicitly skipped (Active_Context is NOT skippable)
+- [ ] Inbox processed or acknowledged as skipped
+- [ ] System improvements discussed
+- [ ] Wins list generated (Step 3G)
+
+If any gate is open: "You haven't confirmed [X]. Wait or proceed?"
+
+---
+
+## HANDOFF
+
+Sources used: [List ALL files read]
+Confidence range: [X-Y]
+
+**Next:** "Weekly Review complete. Ready to run `/plan-tomorrow` for Monday?"
+If Strategy Pivot needed: "First run `/pivot-strategy`, then `/plan-tomorrow`."
+
+Key rules still active:
+- All claims must be grounded
+- Strategy changes require user confirmation
+- Sample size must be met before recommending pivots
+- Append-only: never delete strategy history