# Session Handoff — 2026-07-13 (V1)

**Focus:** GHL Management OS scoping, GHL knowledge gap, memory tool evaluation, "The Floor" internal codename, outreach tooling decisions, lead-magnet rebrand.

---

## Objectives

1. Decide on PowerShell 7 / WSL upgrade paths — answer only, no install.
2. Decide whether to install Hermes Desktop update (0.18.2 → ready version, 257 changes).
3. Diagnose "Update didn't finish — backend update failed" error in Hermes Desktop.
4. Evaluate superpowers (`obra/superpowers`) and other GitHub repos that came up in conversation.
5. Evaluate memory repos: `thedotmack/claude-mem`, `rohitg00/agentmemory`, `MemoriLabs/Memori`, Mem0, ByteRover, etc.
6. Scope "GHL Management OS" — what it actually is, what it isn't, what the real time-sinks are.
7. Push back on premature org-chart / team-naming exercise. Recommend staged build plan.
8. Decide on lead-magnet brand name ("Money Map" → rejected → "The Leak Report" + "Revenue Thief" angle).
9. Draft Apollo.io search filter for Brighton CO pilot.
10. Install Hermes skill: `ghl-expert` to fill GHL knowledge gap.

---

## Decisions Made

### 1. PowerShell 7 + WSL — SKIP for now
- PS 5.1 is fine for current workflow. Side-by-side install (`pwsh.exe`) if ever needed.
- WSL is not needed because VPS 2.25.172.164 is the Linux workhorse. Don't add local Linux env without a real reason.

### 2. Hermes Desktop update — ATTEMPTED
- 0.18.2 with 257 changes available. Clicked "Update now" twice — both failed with "Update didn't finish — Backend update failed."
- Cache-clear attempt did not help. Root cause: Telegram plugin polling conflict, not the updater itself.
- Telegram sessions wedged; no zombie Hermes processes. App remained usable.

### 3. `obra/superpowers` — NOT a fit
- 254k stars, v6.1.1, ~2 weeks old. Real, mature project.
- Purpose: software development methodology for coding agents (Claude Code, Codex, Cursor, etc.) — TDD, spec → plan → subagent-driven implementation.
- Not hermes-aware. Plugin for 10 coding harnesses; Hermes Desktop is not on the list.
- Rob's role is content/marketing/GHL ops, not app development. Skip until work shifts toward building software.

### 4. Memory repo evaluation — DEFER
- Hermes has built-in curated memory (`MEMORY.md` + `USER.md`) + 8 native provider plugins. Use that first.
- `thedotmack/claude-mem` (87k) — coding-agent focused, not hermes-native.
- `rohitg00/agentmemory` (25k) — same, coding-agent focus.
- `MemoriLabs/Memori` (15k) — SQL-native, heavy, enterprise-y. Overkill for personal use.
- Mem0, ByteRover — cloud services, ByteRover has token cost.
- Right time to revisit: when multi-agent setup exists. For now, fix Hermes's current memory (verify `~/.hermes/memory/` is being written to).

### 5. Top GitHub repos by fit for Rob — ranked
🟢 **Actual fit:** `farion1231/cc-switch` (multi-agent switcher), `thedotmack/claude-mem` or `rohitg00/agentmemory` (memory), `kepano/obsidian-skills` (only if Obsidian in use), `NVIDIA/NemoClaw` (sandboxed runtime), `nesquena/hermes-webui` (web/mobile UI), `getpaseo/paseo` (multi-agent orchestration).
🟡 **Conditional:** `ruvnet/ruflo`, `colbymchenry/codegraph`, `garrytan/gbrain`, `MemoriLabs/Memori`, `iOfficeAI/AionUi`.
🔴 **Not for Rob yet:** `NousResearch/hermes-agent` (already using), `mindsdb/mindshub`, `Panel-dev/1Panel`, `screenpipe/screenpipe`, `langbot-app/LangBot`, `jnMetaCode/agency-agents-zh`, `facebook/hermes` (JS engine, not agent).

### 6. "GHL Management OS" reframed — SCRIPTED STAGED PLAN
- Renamed internal codename: **"The Floor"** — the operational layer under Rob. No external branding.
- Rejected the org-chart-first approach. The org emerges from real work, not the other way around.
- Rob = Founder, top of org. Hermes = the manager. Sub-agents for specific jobs come later.
- Sub-agents identified from real time-sinks:
  - **Delivery Agent** (builds what was sold — voice agent, website, workflows, sub-account)
  - **Onboarding Agent** (client checklist, pre-call collection, agenda)
  - **Reporting Agent** (audits clients, computes ROI by category, flags issues)
- Lead gen and the GHL knowledge gap are different problems, not sub-agents.

### 7. Staged build plan — APPROVED
- **Week 1:** Finish 14 outstanding deliverables from launch doc. Run outreach (tomorrow). Get 1 paying client. Install `ghl-expert` skill. **Do NOT build sub-agents yet.**
- **Week 2:** After 1-2 manual builds, turn that template into a checklist (markdown doc, no code). First half of Delivery Agent.
- **Week 3:** Wire GHL webhook so a successful payment fires the checklist. Real sub-agent job #1.
- **Week 4+:** Reporting Agent and Onboarding Agent — only after data exists from real builds.

### 8. Real time-sinks confirmed (3 known, 2+ to be discovered)
1. **Build Voice Booking Agents** — prompts, KB build, 2 dozen voice settings, phone number, webchat bots, calendar link, post-booking workflows.
2. **Build Websites** — GHL Studio AI, SEO/AEO optimization, sitemap, robots.txt, canonical URLs, GA4/GTM, Search Console, GBP connection.
3. **Product + Stripe integration** — recurring packages, 1-time setup fees, "onboarding" trigger, post-payment workflows.
4. (Empty — to be discovered)
5. (Empty — to be discovered)
- Sub-accounts: 1 (Real Results Ready LLC). Pipeline Layer website lives inside that sub-account for now; will get its own sub-account later.

### 9. `ghl-expert` skill — INSTALLED + VERIFIED
- Installed at `C:\Users\Rob\.hermes\skills\ghl-expert\SKILL.md`.
- Skill content: GHL expert consultant persona, plain language, no jargon dumps, prefers native GHL features, web-search to verify uncertain features.
- Verified working: answered A2P approval question (brand reg requirements, common client objections, time expectations, common rejection causes). Test passed.

### 10. Lead-magnet rebrand — "Money Map" REJECTED
- "Money Map" failed the instant-understanding test for first-time readers.
- **Final pick: "The Leak Report"** — short, brandable, pun works for plumbing (biggest vertical), works for HVAC (refrigerant leak), works for roofing (leak). URL `/leak`.
- Backup angle: **"Revenue Thief"** (fear of loss). Also brainstormed "Competitor Pickpocket" (deferred to v2 — needs competitor intel).
- Hero visual: bucket or balloon with dollars pouring out (Rob's idea).
- Page title: "The Leak Report." Headline: "How many jobs did you lose this month?" or "Where's my money going?"

### 11. Apollo + Instantly — Instantly pushed back on
- Apollo free tier (10K credits/mo): install, set billing cap to $0 before first run.
- Instantly.ai ($30-60/mo): Rob wants to install "to be safe" but volume target is 1K/mo pilot, not 5-10K. Launch doc's own rule: "Add Instantly only if bounce > 3% or spam folder placement." Recommend waiting for data. Rob may proceed anyway — his call.

### 12. Apollo search filter — DRAFTED (not yet executed by Rob)
- Boolean for Brighton CO + 10 miles, 4 verticals, 5-50 employees, decision makers.
- Exclusions: franchises, 100+ employees, national brands (ServiceMaster, Roto-Rooter, Mr. Rooter, Benjamin Franklin).
- Title filter: Owner, General Manager, Office Manager, Dispatcher, Marketing Director.
- Seniority: owner, c-level, director, manager.
- "Email verified" filter ON.
- Billing cap to $0 (or $49 max) before first run.

---

## Configuration Changes

- **Hermes Desktop:** update attempted twice, both failed. App still on 0.18.2. Telegram polling conflict is the underlying cause. No config fix attempted; recommendation is to wait for the wedged session to expire (5-10 min) and retry.
- **Hermes skill:** `ghl-expert` installed at `C:\Users\Rob\.hermes\skills\ghl-expert\SKILL.md`. Restart Hermes Desktop to load.
- **Apollo.io:** account creation in progress (Rob setting up tonight).
- **Instantly.ai:** Rob may set up tonight despite pushback.

---

## Files Edited

- **Created:** `C:\Users\Rob\.hermes\skills\ghl-expert\SKILL.md` (new skill, see Configuration Changes above).

No other files edited. No GHL config, no GHL workflows, no Apollo config, no Instantly config, no site pages, no Money Map / Leak Report pages.

---

## Outstanding Issues

### Blockers / Decisions still pending Rob

1. **GHL update retry:** Wait for Telegram session to expire (5-10 min from last conflict log) and click "Try again" on the updater. If still failing, do clean restart of Hermes Desktop.
2. **Apollo billing cap:** MUST be set to $0 (or hard cap) before first search/run. Silent credit reload is a known trap.
3. **Instantly.ai:** Decide whether to install now or wait for first deliverability issue. Recommended wait; Rob leaning install.
4. **The 14 deliverables from launch doc** (see Next Steps).
5. **GHL custom fields (13 total):** confirm 4 existing fields from outreach skill still apply + 9 new fields for the Brighton CO pilot funnel. Not done in this session.
6. **Sending domain DNS:** reach.realresultsready.com vs new domain. Not decided.
7. **Postcard printer (Postable vs Lob):** Not decided in this session.
8. **Lead-magnet form fields:** 4 main + 2 scoring (business size, call volume bucket). Not finalized.

### Knowledge gaps (deferred to future sessions)

1. Brighton CO 10-mile radius Apollo boolean for 4 verticals — drafted tonight, needs Apollo UI testing.
2. Postable vs Lob GHL integration paths (Zapier vs API).
3. USPS CASS/NCOA processing — bundled or separate.
4. Brighton CO contractor license lookup for lead sourcing.
5. GHL custom-page form calculation logic (real-time math vs custom HTML/JS).
6. GHL native cold email deliverability benchmarks (LC Email).
7. SPF/DKIM/DMARC setup for reach.realresultsready.com — exact DNS records.

### Token security

- Rob confirmed: **no GHL API token was pasted in chat.** (Learned the habit earlier in the day.)
- If a token ever IS pasted, rotate immediately at `app.gohighlevel.com → Settings → API`.
- General rule: never paste secrets/tokens in chat. Save to local file on disk, reference the path.

---

## Next Steps

### Tonight (Rob's homework)

1. ✅ Set up Apollo.io free account.
2. ⚠️ Set up Instantly.ai (or defer — Rob's call).
3. ✅ Paste the Brighton CO boolean into Apollo (saved in Decisions #12 above). Verify it returns a reasonable result set.
4. ✅ Set Apollo billing cap to $0/mo before first run.
5. 🔲 Knock out 2-3 of the 14 launch-doc deliverables. Recommended order:
   - #2 / #13 (Apollo filter) — already drafted, just paste + verify.
   - #3 (3-touch GHL follow-up sequence) — 30 min.
   - #1 (HVAC vertical landing page outline) — 30 min, outline only.
   - #6 (ROI math per vertical) — 20 min, numbers already in pricing doc.
6. 🔲 Finalize lead-magnet rebrand: confirm "The Leak Report" as page name. Begin landing page copy when ready.
7. 🔲 Optional: re-attempt Hermes Desktop update after waiting for Telegram session to expire.

### Tomorrow (Tuesday)

1. **Send cold email outreach** — 1K email pilot in one vertical/geo first (per launch doc).
2. **Send Brighton CO postcard pilot** (if printer account ready and list acquired).
3. **Set GHL custom fields** (4 existing + 9 new = 13 total) so form submissions can flow into the right places.

### This week (Week 1 of staged plan)

1. Finish the 14 deliverables.
2. Run outreach.
3. Get 1 paying client.
4. **Do NOT build sub-agents yet.** Use the ghl-expert skill to fill GHL knowledge gaps.
5. Manually build 1 client sub-account end-to-end. Capture the steps as a checklist.

### Next week (Week 2)

1. Convert the manual build into a markdown template.
2. Add the ghl-expert skill reference to the template.
3. No code yet — just a checklist that any future sub-agent could follow.

### Week 3

1. Wire the GHL webhook: successful payment → checklist fires → Rob gets notified.
2. Delivery Agent v1 (real sub-agent job #1).

### Week 4+

1. Reporting Agent (after 2+ real builds have data).
2. Onboarding Agent (after manual onboarding has been done 2+ times).
3. Lead-gen sub-agents (separate problem, not started yet).

### Long-term / Parked

- Pipeline Layer workshop/scorecard funnel (referral + thought leadership track, NOT cold email).
- LinkedIn 1:1 outbound program (50 targeted CMOs/RevOps leaders/week).
- Content engine (1 piece of thought leadership bait per week).
- "The Floor" org chart (only as new sub-agents are justified by real work).

---

## Context for Next Session

### Key data points

- Apollo.io free tier: 10K email credits/mo, 60 mobile/yr, 120 export credits/yr.
- Real Results Ready avg customer LTV: $5,400 (18-month avg × $300/mo).
- Real Results Ready pricing tiers: Capture $497/mo, Convert $2,497/mo (most clients), Compound $4,997/mo.
- Pipeline Layer (separate brand): $7,550 AI Audit — referral + thought leadership, NOT cold email.
- Brighton CO + 10-mile radius: 180-300 local service businesses. ZIPs 80601, 80602, 80603.
- Avg job value per vertical: HVAC $500-2K, Plumbing $300-1.5K, Electrical $200-800, Roofing $5-15K.
- Missed call rate in home services: 14% (CallRail 2025 benchmark).
- Direct mail conversion: 2-5% QR scan, 1-4% form completion, 0.5-2% new customer.
- GHL tier: $297/mo (agency), confirmed.

### Confirmed preferences

- Low-context delivery: action + minimum context, no recap.
- Hates "wrong half the time" — wants complete fixes.
- Cost-conscious, wants API cost visibility + monthly caps.
- Zero tolerance for silent cost leaks.
- Style: casual lowercase, abbreviated.
- Output format: PowerShell/batch copy-paste blocks get "Copy and paste this whole block" header + closing separator. Reference material gets "for your records" label.
- SECURITY: when a secret/token/key will be pasted, EXPLICITLY warn "do not paste the actual value in chat" UPFRONT in the FIRST relevant block, not buried at the end.

### GHL-aware skill now active

- `ghl-expert` — load by asking any GHL question. Test passed with A2P approval deep-dive.

### Session scoreboard

- Decisions made: 12
- Files edited: 1 (ghl-expert skill)
- Config changes: 1 (skill install, plus attempted Hermes Desktop update)
- Open questions for Rob: 8
- Next session priority: confirm Apollo filter results, decide on Instantly, knock out 2-3 of 14 launch-doc deliverables, finalize "The Leak Report" landing page copy when ready.

---

*End of handoff.*