Formalize BACKUP BEFORE UPDATE rule (2026-02-06)

- Added explicit rule to MEMORY.md Automated Maintenance Rules
- Added rule to fleet-wide OPERATIONAL-GUIDELINES.md (top priority)
- Added Lesson vignesh07#34 about this incident
- Updated daily memory with incident details

Author: Molty

MEMORY.md

@@ -593,6 +593,22 @@ bash scripts/api-capture/capture-and-generate.sh example.com --timeout 120
 
 33. **Formalizing sub-agents pays off** — Ad-hoc spawning works but named roles with pre-loaded instructions = faster + more consistent results.
 
+### Day 6 (2026-02-06)
+
+34. **BACKUP BEFORE UPDATE — ALWAYS** — System crash reinforced this rule. The procedure was documented but as "after backup → check updates" not "before update → backup first". Made the rule explicit and bidirectional. Added to fleet-wide OPERATIONAL-GUIDELINES.md. No exceptions. (2026-02-06)
+
+---
+
+## 🔄 Automated Maintenance Rules
+
+| Trigger | Action | Notes |
+|---------|--------|-------|
+| **BEFORE any update** | **Backup first** | **NON-NEGOTIABLE** — run backup.sh before git pull/pnpm install/restart |
+| After backup (every 6h) | Check for OpenClaw updates | Run `git fetch` + review commits, apply if safe |
+| After update | Notify Guillermo via active channel | Confirm version + restart |
+
+⚠️ **BACKUP BEFORE UPDATE** — This rule was set 2026-02-05 and must be followed by ALL agents. Guillermo explicitly mandated this after a system crash on 2026-02-06. The backup takes ~2 minutes and can save hours of recovery work.
+
 ---
 
 ## 🔧 Quick Reference

memory/2026-02-06.md

@@ -0,0 +1,42 @@
+# 2026-02-06 (Thursday → Friday)
+
+## Automated Updates
+
+### 02:48 UTC (10:48 HKT)
+- **Raphael updated** via scheduled cron job
+- Commit: 7b2a2212 (origin/main)
+- Backup: raphael-backup-20260206-024616.tar.gz (2.4M)
+- Build: pnpm install + pnpm build succeeded
+- Restart: SIGUSR1 → gateway restarted
+- Status: ✅ Up to date, reachable
+
+Security note acknowledged: `dangerouslyDisableDeviceAuth=true` flagged but is known config (both instances).
+
+---
+
+## System Crash & Rule Reinforcement
+
+### ~03:30 UTC (11:30 HKT)
+- **System crash** on lucky-bl (SIGKILL)
+- Guillermo requested:
+  1. ✅ Update OpenClaw
+  2. ✅ Tell Raphael to backup then update
+  3. ✅ Check what rule was forgotten about backups
+
+**Root cause of confusion:** The maintenance rule was documented as "After backup → check for updates" but NOT explicitly as "Before update → backup first". The logic was there but implicit.
+
+**Resolution:**
+- Updated MEMORY.md "Automated Maintenance Rules" section to make "BACKUP BEFORE UPDATE" explicit and NON-NEGOTIABLE
+- Added to fleet-wide OPERATIONAL-GUIDELINES.md as top-priority rule
+- Added Lesson #34: "BACKUP BEFORE UPDATE — ALWAYS"
+- Sent webhook to Raphael acknowledging the formalized rule
+
+**Note:** Raphael actually DID backup before his update (the log shows it). The rule was followed in practice, just not documented explicitly enough for autonomy.
+
+### Molty Update Applied
+- git pull + pnpm install + pnpm build succeeded
+- Gateway restart triggered
+- Now at latest commit
+
+---
+