---
title: vault/ Migration Audit — Deep Dive
created: 2026-05-01
type: investigation
status: read-only
author: claude-code-rr (4 parallel Explore agents)
---

# vault/ Migration Audit — Deep Dive

## TL;DR

Four parallel agents read every meaningful file in the off-limits `vault/` directory (1,796 files total, classified across 11 top-level folders). Big picture:

- **~28 files (~530 KB) belong in HQ-Vault.** Operational content — business model, competitors, roadmap, action queue, project state, recent dailies, agent outputs.
- **~30 files in `04-Operations/` are reusable SOPs** (deployment, incident response, email sending, agent recovery, model change). Worth pulling into `90_Reference/` or `11_OpsMan_LFCS/`.
- **~1,731 files (~40 MB) stay in `vault/09-Agent-Archive/`** — pre-hibernation agent state for Susan, Harper, Rivet, Herald, Radar, Sentinel, Cog, Builder. Rocky said "might bring them back". Keep as reference.
- **Secrets audit: clean.** No API keys, passwords, full bot tokens, full card numbers in any of 1,796 markdown files. Compliance maintained.

But the **bigger finding** isn't the migration list — it's what's IN those files.

## The bombshell — your own words from 2026-03-30

`vault/01-Daily/2026-03-30-System-Audit.md` is a self-audit you did 32 days ago that identifies **exactly the problems we've been fixing this week**:

> *"Michael's decisions are not reaching the fleet through the vault. Example: ADI bank account blocker said 10+ times, never recorded as a hard constraint, Harper looped chasing the dead grant for 30 days."*
>
> *"Code disconnect: QR funnel built/tested locally on Michael's laptop, VPS fleet never received the code. Builder idle."*
>
> *"Memory black hole: Michael said 'X' ten times, fleet failed to record it as a hard constraint."*

You called out **the exact failure mode the AI Team Protocol was designed to prevent**: decisions not propagating, code not syncing, fleet paralysis from communication breaks.

## Action items from 2026-03-30 that are still pending today

`vault/07-Command-Centre/Michael-Action-Queue.md` (last updated 2026-03-30, **5 weeks ago**):

- [ ] **Push local code (QR funnel fix) to VPS** — still pending today
- [ ] **Drop finalized hostel poster PDF into vault for mailouts** — still pending today
- [ ] **Hook up Google Workspace CLI (gws) on VPS** — still pending today

Plus from `06-Business/Revenue Pipeline.md`:

- [ ] **Pawel Wolonciej (AWX Labour Hire)** — 40+ days overdue for callback as of 2026-03-12. Now ~80 days overdue.
- [ ] **9 warm leads uncontacted** — Susan blocked on your approval. Approval gate set 2026-02-17. Stale ~75 days.
- [ ] **17 lead-scored 65-85 + 10 SMS drafts staged ready** in Susan's MEMORY.md — never sent.

## Other things you should know

### Grant officially dead 2026-03-30
**MVP Ventures grant** ($75K, deadline 2026-04-05) was officially cancelled on 2026-03-30 because Airwallex isn't ADI-compliant and you needed an ADI bank account. Deadline has now passed; grant is dead.

But: `vault/07-Command-Centre/Roadmap.md` (last updated 2026-04-09, **9 days AFTER you cancelled the grant**) still has Harper "actively pursuing alternatives". The decision didn't propagate to the roadmap. Same problem you identified.

### Fleet was already paralysed by 2026-04-09
The April 9 roadmap shows: **entire fleet offline 21+ hours, no heartbeats. API budget exhausted. Sentinel, Cog, Radar offline. Builder offline (bridge connection refused). Susan stuck in error loop.** That document is 22 days old. The hibernation on 2026-04-29 was effectively the formal acknowledgement that the fleet had already collapsed.

### What Susan had ready to ship (still live in `09-Agent-Archive/Susan/MEMORY.md`)
- **17 contractor leads scored 65-85** with phone + email ready
- **10 personalised SMS drafts** staged for approval
- **Refined value prop v4**: "$50. One-off. Find a rated [trade], hire directly. They stay a week or 30 years — still $50."
- **Live competitor intel** (Feb 2026): hipages losing 35,000 subscribers, MRR retention 58%, ACCC targeting their auto-renew. Airtasker $16.5M H1 loss. Sidekicker now a SEEK property.
- **Sales playbook**: target 5-500 employee companies, Tue-Thu 9-11am, 4 objection scripts.
- **Sydney market intel**: 31K homes in TOD program, $7B pumped hydro, 16 school upgrades — every builder fighting for same tradies.

This is **months of polished, thought-through, ready-to-ship sales material** sitting in an archived agent's workspace. Worth either reviving Susan or extracting the playbook.

### What Harper had ready
- **All 6 MVP Ventures grant docs drafted** (now moot)
- **8 tax-optimisation strategies identified** (~$8-9K/year savings)
- **Insurance audit**: PI + PL needed (~$1-2K/yr) before launch
- **National expansion model**: 7-city launch at $723/month marketing, breakeven Month 7-14
- **Hostel direct mail business case**: $14/hostel all-in, 148 hostels mapped, CAC $3-7/signup (5-10× cheaper than digital ads)

### What Radar had researched
- **23 unique hostel emails** across Melbourne, Perth, Darwin, Hobart, Canberra (Tier 1 = already running construction job boards)
- **24 Sydney hostel emails** across CBD, Surry Hills, Bondi, Manly, Kings Cross, Glebe, Newtown, Coogee — **44% coverage of ~54 Sydney hostels**

### What Builder had built
- Smart Matching Algorithm verified
- Phone collection in all signup flows
- Seed data: 10 workers, 3 contractors, 5 jobs
- 4 critical bugs fixed (JS chunk loading, server actions, manifest, MIME)
- Build passes, app demo-ready
- **Stripe integration tested, not live** (still the critical gap to launch)

---

## Migration plan — what should move to HQ-Vault

### Tier 1 — Migrate as-is (high signal, current)

| File (vault/) | Move to | Why |
|---|---|---|
| `06-Business/Business Model.md` | `10_RateRight/_Brain/Business-Model.md` | Core $50 thesis, unit economics, current numbers (304 contractors, 0 workers, $24,408 cash, 31mo runway) |
| `06-Business/Competitors.md` | `10_RateRight/Business/Competitors.md` | Threat matrix (SEEK/Sidekicker, Yakka, Casu, Juicebox) |
| `06-Business/Revenue Pipeline.md` | `10_RateRight/Business/Revenue-Pipeline.md` | Lead status (Pawel, Tony McCabe, etc) — needs status header noting staleness |
| `07-Command-Centre/Decision Log.md` | `10_RateRight/_Brain/Decision-Log.md` | Architecture decisions, fallback chain, Obsidian SSOT — append-only style worth keeping |
| `07-Command-Centre/Operations Architecture.md` | `11_OpsMan_LFCS/_Brain/Operations-Architecture.md` | Three-layer sync model documentation — actually correct, just stale at the implementation layer |
| `07-Command-Centre/Daily Workflow.md` | `_System/Daily-Workflow.md` | Your schedule (4:15 gym → site → Window → coding) — already in `_System/CLAUDE.md` Block A but full version is here |
| `07-Command-Centre/Daily-Team-Audit.md` | `11_OpsMan_LFCS/Operations/Daily-Team-Audit.md` | The 10-min morning protocol — solid SOP |
| `03-Projects/The $50 App.md` | `10_RateRight/_Brain/Platform.md` | $50 App architecture (Next.js, Supabase, Stripe, hosting) |
| `03-Projects/Growth Engine.md` | `10_RateRight/Codebase/Growth-Engine.md` | Backend architecture |
| `03-Projects/Control Centre.md` | `11_OpsMan_LFCS/Codebase/Control-Centre.md` | Fleet UI architecture |
| `03-Projects/Voice AI.md` | `11_OpsMan_LFCS/_Brain/Voice-AI.md` | VAPI + Pipecat docs |
| `03-Projects/Worker-Contractor-Acquisition.md` | `10_RateRight/Business/Worker-Contractor-Acquisition.md` | Strategy doc — needs status header noting hostel mailing stalled |
| `09-Agent-Output/2026-03-13_radar_hostel-emails-melbourne-perth-rest.md` | `10_RateRight/Business/Marketing/Hostel-Emails-Melbourne-Perth.md` | 23 emails, ready to send |
| `09-Agent-Output/2026-03-13_susan_sydney-hostel-emails.md` | `10_RateRight/Business/Marketing/Hostel-Emails-Sydney.md` | 24 emails, ready to send |
| `09-Agent-Output/contractor-cold-email-draft.md` | `10_RateRight/Business/Marketing/Contractor-Cold-Email-Draft.md` | Approved draft, ready |
| `01-Daily/2026-03-12.md` | `01_Daily/2026-03-12.md` | OpenRouter incident response + agent archive consolidation |
| `01-Daily/2026-03-25-Grill-Session.md` | `01_Daily/2026-03-25.md` | **Brutal accountability session — must read** |
| `01-Daily/2026-03-30-System-Audit.md` | `01_Daily/2026-03-30.md` | **Surgical root-cause analysis — must read** |

### Tier 2 — Migrate with status header (high signal, dated)

| File | Move to | Status header to add |
|---|---|---|
| `07-Command-Centre/Roadmap.md` (2026-04-09) | `10_RateRight/ROADMAP.md` (REPLACES the older 2026-03-16 version) | "Status as of 2026-05-01: Hibernation occurred 2026-04-29. Fleet retired. Hermes + OpsMan + Rivet Voice survive. Hunt scrapers blocked on ScrapingBee credit until 2026-05-13. Action items below were rendered moot by hibernation; reset priorities in The Window." |
| `07-Command-Centre/Dashboard.md` | `10_RateRight/_Brain/Dashboard.md` | "Reflects state at 2026-03-12. Most numbers stale; treat as historical baseline. See HealthLog and live VPS state for current." |
| `07-Command-Centre/Critical Issues.md` | `10_RateRight/_Brain/Critical-Issues-2026-03.md` | "Snapshot of issues circa 2026-03-12. P0/P1 items have been resolved or superseded. Kept for historical pattern recognition." |
| `07-Command-Centre/Michael-Action-Queue.md` (2026-03-30) | `10_RateRight/_Brain/Michael-Action-Queue.md` | "Items below are 5+ weeks old as of 2026-05-01. Some still pending (push local code, drop poster, gws on VPS). Review in The Window and either action or move to _Archive." |

### Tier 3 — SOPs to migrate to `90_Reference/` or `11_OpsMan_LFCS/`

| File | Move to | Why |
|---|---|---|
| `04-Operations/SOPs/Deployment.md` | `11_OpsMan_LFCS/Operations/Deployment-SOP.md` | Next.js build-restart pairing rule, Railway deploy pattern. Reusable. |
| `04-Operations/SOPs/Incident Response.md` | `11_OpsMan_LFCS/Operations/Incident-Response-SOP.md` | P0/P1/P2 severity matrix, "diagnose first" rule from Feb-18 incident. |
| `04-Operations/SOPs/Email Sending.md` | `11_OpsMan_LFCS/Operations/Email-Sending-SOP.md` | Gmail API CLI usage, OAuth pattern, 2,000/day limit. Active. |
| `04-Operations/SOPs/Agent Recovery.md` | `11_OpsMan_LFCS/Operations/Agent-Recovery-SOP.md` | systemd recovery + JSON config validate + session override clear. |
| `04-Operations/SOPs/Model Change.md` | `11_OpsMan_LFCS/Operations/Model-Change-SOP.md` | Config trap (`/root/.openclaw/` not `/root/.clawdbot/`), session override cleanup. Saved me yesterday. |
| `04-Operations/Comms System Investigation.md` (2026-03-17) | `_System/Investigations/Comms-System-Investigation-2026-03.md` | Full architecture deep dive of HQ↔COO↔CC-Local↔CC-VPS↔Rivet↔Fleet. Patterns reusable for current Hermes-OpsMan setup. |
| `Templates/Daily Note.md`, `Incident Report.md`, `Brain Dump.md` | `_System/Templates/` (only if new) | Compare with existing; replace if better. |

### Tier 4 — Stays in `vault/` (per your "might bring back")

- **All of `09-Agent-Archive/`** (1,731 files): Susan's 172-file workspace including the 284-line MEMORY.md sales playbook. Harper's 117 files. Rivet's 506 files including LESSONS.md (12 hard-won incidents). Herald's 281 files including 170+ content drafts and 16-city hostel poster library. Radar's 63 research files. Builder's 17 files including the seed data + bug audit. Etc.
- **`02-Fleet/`** (10 agent profile cards + Fleet Map) — old architecture reference if you revive
- **`05-Knowledge/Fleet Model Plan.md`** — old fleet model strategy
- **`08-Infrastructure/`** (4 files) — old infrastructure topology

### Tier 5 — Don't migrate (superseded / dead)

| File | Why |
|---|---|
| `06-Business/Grant Tracker.md` | Grant cancelled 2026-03-30. Dead. (But preserve in vault for the record.) |
| `04-Operations/Incidents/*` (3 files) | Pre-hibernation outage records. Historical. Stay. |
| `vault/Templates/*` if HQ-Vault templates are equivalent or better | Compare before deciding. |

## What this audit reveals about how to use the team

The 2026-03-30 audit identified the failure mode. The AI Team Protocol I committed today addresses it directly:

| 2026-03-30 problem | AI Team Protocol fix |
|---|---|
| Decisions don't reach fleet through vault | Boot sequence requires reading `_Brain/CLAUDE.md` + `_Log.md` every session |
| Memory black hole (said "X" ten times, fleet didn't record) | Append-only `_Log.md` + `01_Daily/` enforced via pre-commit hook (`scripts/check-append-only.sh`) |
| Code disconnect (built locally, never synced to VPS) | Auto-sync every 30 min in both directions; vault-verify gate before push |
| Fleet paralysis from API budget | Per-agent fallback chains documented; Hermes already has `MiniMax → DeepSeek → gpt-4o-mini` |
| Grant cancelled but roadmap still chased | Append-only history means a contradicting later decision lands as a new entry, not a silent rewrite |

So the failure mode you flagged 32 days ago is the SAME failure mode I've been wiring guardrails against this week. The vault you've already got documents the disease; the protocol we just put in place is the cure.

## Sensitive data audit

Across all 1,796 files: **clean.** No API keys, passwords, full bot tokens, full card numbers, or other credentials in any markdown body. References by name only. Auth profile **locations** documented but credential **values** are externalised to `.env` / `secrets.json` / `/root/.openclaw-{agent}/auth-profiles.json` as expected.

One minor exposure: ElevenLabs voice ID `IKne3meq5aSn9XLyUdCD` is in `vault/03-Projects/Voice AI.md` and `08-Infrastructure/Voice Systems.md`. This is a public identifier (their voice catalog) not a secret; low concern.

## Recommended next moves

Three options, in increasing scope:

1. **Surgical (default — what I'd do)**: Migrate Tier 1 + Tier 2 only. ~22 files. One commit per logical group. ~30 minutes work for me. Leaves SOPs and stays-in-vault content untouched. You then have the operational content where the AI Team Protocol expects it.

2. **Surgical + SOPs**: Tier 1 + Tier 2 + Tier 3. ~30 files. Adds 60 minutes for SOP migration with status notes. Gives you reusable SOPs in HQ-Vault for future reference.

3. **Full operational migration**: Everything except Tier 4 + Tier 5. ~32 files. Full reorganisation pass.

After any of these, `vault/` becomes ~98% pre-hibernation agent archive — clean separation from your active brain.

I'd default to **option 1**. Tell me to go and I'll execute. Or pick option 2 / 3.

## Read-only

This audit made no changes to either vault. Migration is awaiting your call.