1 Today at a glance
- Lead hand-off is fixed, proven, and now instant. This morning a test lead ran the whole path (tagged qualified → queued → texted) with phone-carrier confirmation. This afternoon the delay was cut from up-to-5-minutes to ~5–11 seconds, proven again with live tests. Last step: one real outside person runs it end-to-end (your own numbers are blocked by design).
- The Texas expansion plan is written and published. Houston + Austin launch is fully scoped on one page — it needs ~15 minutes of your Facebook/CRM clicks and 4 decisions, then agents build everything (paused) for your approval.
- Two new Facebook pages are live: Texas statewide + Austin concrete-leveling pages (one page per state is the new scaling play — ~15× faster than per-city pages).
- The competitor swipe file is now serious: ~2,800 real ads collected today — the biggest players in foundation repair and basement work, plus every direct concrete-leveling competitor found — with 900+ ad videos safely backed up. Market research also confirmed: nobody in concrete leveling sells leads or pay-per-result — you'd be a category of one.
- The 30-day focus is locked and the task board is live. 96 unrelated tasks parked until Aug 27; Linear has 28 active plain-English cards with two-way sync.
- The Mac got 9.6 GB lighter. Cold video files were uploaded to Drive, verified byte-for-byte, then removed locally. Nothing was deleted unverified.
- Watch-out: both weekend check-ins silently stalled. The weekly cash check and the Sunday projections never actually ran — they froze on startup and sat "running" for 24+ hours. The weekly cash number and its safety tripwires were NOT checked this week (the Jul-31 no-signed-client gate is 3 days out). Both get re-run and diagnosed — see section 5.
- The books are clean: all 15 of today's sessions are saved to the brain and every transcript is backed up (1,658 of 1,658 in the archive).
2 Your links — everything in one place
Start here
Scoreboards
Fresh today — needs your eyes
Reference docs
| What | Where |
| Operating playbook (how the whole business runs) | Documents/Claude/02-Areas/Team/Internal Docs/operating-playbook/index.html |
| Lead outcome definitions (what counts as qualified, transferred, won) | 01-Projects/St_Pierre_Digital/Internal-Docs/qualified-conversation-outcomes-2026-07-28.md |
| CRM tag audit (what tags actually exist vs what they should be) | 01-Projects/St_Pierre_Digital/Internal-Docs/ghl-tag-audit-2026-07-28.md |
| Task-board build plan (phases still open: feel test, old-tool retirement, agent identity) | stpierre-linear-workspace-sow.pages.dev |
3 The plan — one SOW, four steps
The goal: concrete-leveling leads flowing to Dan every day — 25 hand-offs a week. Everything below serves that; everything else is parked until Aug 27.
| Step | Outcome that proves it | Detailed plan | Who |
| 1 · Prove delivery (this week) | A real customer message reaches Dan's phone in seconds, three days straight. Delivery is already instant (~5–11 s, proven today) — what's left is the bot hand-off text, the follow-up cadence, and one outside tester. | Lead Machine plan | You + agents |
| 2 · Scale volume | Ad sheet live, Dallas budget confirmed, Texas + Austin open for business, next markets minted | Texas launch plan | You approve, agents build |
| 3 · Keep it healthy | Daily ad batch, fresh reply drafts, green morning health report — without you touching anything | Section 5 below | Agents |
| 4 · Package it | Dan's setup becomes a reusable checklist so client #2 onboards in days, not weeks | Section 6 below | Collins (gated) |
Sections 4–6 below are the complete task lists for each step, with everything each person needs. The two plan pages carry the full context — paste-ready text, click lists, and the decisions each step is waiting on.
4 Waiting on you — calls, clicks, approvals
Phone & messages (revenue today)
- Send Dan the two finished lead packets (Sam and Spencer — formatted and ready to paste; nothing was sent automatically).
- Call the leads that have sat cold: Cindy, Richard, Ann — plus Cindie wants work done next week, and Ann + Scott are owed apologetic follow-ups. Reply to the three unanswered Level King leads.
- Send the four ready-to-go follow-up drafts before the reply window closes tomorrow (Jul 29).
- One text to Dan, four asks: his follow-up-software key · his exact service area · his booking-form link · confirm he's good with the new instant delivery. This unblocks delivery into his system and widening ads to Fort Worth / Arlington / Grand Prairie.
CRM clicks (guides + paste-ready text on the Lead Machine plan)
- Paste the ready-made hand-off instruction into the bot (the 496-character block is on the plan page, copy-paste).
- Set up the 6-step follow-up cadence (numbered click list on the plan page).
- Flip the cleanup toggles: 6 safe automation drafts + 3 clicks from today's automation audit (pause the pre-call nurture, delete the dead auto-transfer automation that has never fired, disable the duplicate ad-event step).
- The audit pack still has the morning list: confirm the "qualified" tag step, add the phone+ZIP gate, clear 4 stale review tags.
Facebook clicks (~15 min — this exact step has blocked Houston/San Antonio since yesterday)
- Claim the new TX and ATX pages into the business account.
- Grant the automation user access to all four concrete-leveling pages — after this, agents dress the pages (photos, details) automatically.
- Connect the Texas page in the CRM's Facebook settings so the bot answers its Messenger inbox.
Decisions (one sitting)
- 4 Texas decisions on the launch plan: daily budget ($50/day suggested) · Dallas stays on its current campaign · which ad account builds it · the statewide bot's name.
- Review + activate the concrete-leveling ad sheet (built, paused).
- Confirm the Dallas campaign's budget intent — it's live at $100/day but still named "$50/day".
- Pick the delivery path into Dan's job software: bare cards now, test his API key, or wait for the connector.
- Line up one outside person (not you, not Dan) to message the page as a fake customer — the final proof of the whole lead path.
- Storage: OK a Drive-space bump (only ~13 GB headroom left) or prune old archives · glance at the 46 MB "New Folder With Items" in Downloads · decide on the ~70 GB staged macOS update.
- The standing queue: remove the "I'm a real person" line from client bots · landing-page form fixes · turn on call recording in the CRM · decide cold-email capacity · lift the manual-send hold when ready · submit the Facebook app review before the trial deadline.
5 Running on autopilot — agents handle these
- New today: re-run the two stalled weekend check-ins (weekly cash check + Sunday projections) and diagnose why scheduled runs freeze on startup before the next ones fire. The cash tripwire check comes first.
- New today: fix the 5 open items from the lead audit (a duplicate contact, two deleted contacts to re-create, debug text to scrub from packet notes, dead follow-up nudges) + build durable photo pages for the three leads with media.
- In progress: 3-day watch on the Dallas client bot — proving one real lead runs every step and reaches Dan in seconds.
- Daily ad batch for each active client + our own account — staged paused with screenshots for your OK, never self-launching.
- Reply drafts kept current for every live lead (sending stays manual until you lift the hold).
- Morning health report checked; anything red gets fixed and logged with an undo.
- Client messages cleared and check-ins hit daily.
- Next market pages minted as Facebook's daily limit allows: Oklahoma City next, then Tulsa.
- New ad creative wave + bot qualification tune (~55% → 70%) staged as approval packs. Sort the new swipe-file ads into ours-vs-clients buckets (~15 min, feeds the creative wave).
- Page dress-up (photos/details) for all four pages — blocked until you do the Facebook claim + assign clicks above.
- Houston/Austin ad build — blocked until the Facebook clicks + 4 Texas decisions above.
6 Collins' build list — gated until you say go
Four cards are labeled for Collins on the task board. Nothing gets sent to him without your explicit go (scope + pay + go, per your rule).
- Map today's client-setup steps — what's built vs missing
- Turn Dan's setup into a reusable checklist for the next client
- Design the setter tracking + coaching system
- Build the daily coaching brief for each setter
Heads-up from today's audit: these cards are not yet send-ready. Each has a one-line goal but no links, no access notes, and no background — Collins can't see your internal notes, so he couldn't finish any of them from the card alone. Before sending, each needs a context pack: relevant links (playbook, boards, call-data source), what access to request, and the definition of done. Say the word and I'll draft all four packs for your review.
7 Board health check — your three questions, audited
| Question | Verdict | Detail |
| Plain English, jargon-free? |
27 / 32 pass |
All 28 active cards read clean; code names live only in collapsed technical notes. Fails: 1 old technical card (the Facebook-login checkpoint one, parked) + Linear's own 4 sample onboarding cards — worth archiving. |
| Does each card link its notes, deliverables, SOW? |
No |
No card links any deliverable or plan doc. Every card has a one-line definition of done + an internal reference id. AI agents can follow that id into the brain; a human can't. Top ~10 cards should get 2–3 links each (this page makes that easy). |
| Could Collins execute his cards alone? |
Not yet |
All four Collins cards fail the hand-off test (see section 6). Fix = context packs, ~30 min of work, ready on your go. |
| Are AI-ready cards complete? |
Pass, with caveat |
7 agent-ready cards all carry a definition of done + brain reference id, and agents resolve full context through that id. Caveat: the card alone isn't enough without brain access — fine for our agents, but add links if outside tools ever pick them up. |
| Everything non-Dan parked? |
Mostly |
96 tasks parked and hidden from every surface (verified zero leaks). Borderline calls still active: "get our own ads running again" + "ship our own daily ad batch" (that's marketing your agency, not Dan's leads) and the two landing-page form approvals. Your call: park them to Aug 27 or confirm they stay. |
8 Cleanups & technical notes
- The stale brain copy in the home folder is fixed for good — all seven scripts that kept regenerating it were patched today and verified.
- Mac cleanup swept 9.6 GB of cold video to Drive (verified byte-for-byte before anything was removed).
- Still queued (15 min): commit today's uncommitted code changes, and lock down the client report link before client #2 onboards.
Technical notes (for agents — safe to ignore)
- Handoff rows written today: 2004–2016 (morning + midday sweeps) and 2017–2023 (evening shutdown: SFFR ad scrape, media sweep, Texas CL SOW, instant-send session, 3 zombie loop sessions). All carry real harness session ids (local_<uuid>).
- Zombie loop sessions: sp-loop-sunday-projections (local_653b4781), sp-daily-pack-pilot (local_e84cea01), monday-cash-check (local_bb2aaf99) all froze on their first Bash calls ~15:33 UTC Jul-27 — likely a permission gate in non-interactive runs. Diagnose before their next scheduled fires; re-run cash check + projections manually.
- Instant-send: DB trigger + webhook on the n8n drain, proven 4.6 s queued→sent and 11 s tag→SMS carrier-verified. AUTOSEND stays false; n8n drain remains THE sender. Open mystery: the GHL workflow Akash says he published is not visible via API in the SP location (125 workflows unchanged since Jul-25) — the 11 s test succeeded through the OLD live workflow, so delivery is instant either way.
- GHL workflow-builder iframe silently swallows agent clicks (control-tested today, zero writes made) — UI builds remain Akash-only, per standing rule.
- Meta write path: system-user token CAN write to the CL client ad account (act_3289323034532141); B2B block unchanged. New drift (high): no stored Meta token can see the CL Dallas ad account — Texas campaign build needs re-auth or a browser-driven build.
- Ad-library skill drift: gws drive files list uses --params not --json; strip the "Using keyring backend" line before parsing gws JSON.
- Texas bot decision: per-ad-set bots REJECTED — one market-aware bot statewide (bot collects ZIP); San Antonio skipped, Houston + Austin only.
- Linear sync: echo guard disabled until agent OAuth identity (P7); receiver workflow 3xeSHn6EYegPjG6a; map-aware upsert fixed duplicate rows.
- Rule reaffirmed today: no transfer-rail fix counts until an end-to-end synthetic-lead test with carrier delivery proof passes.