System

System and Audit

How everything connects, how healthy each part is, what is missing and what to do first. Every line says if it was checked live or comes from our docs.

✎ Edit the text version

Flowchart

Drawing…

Summary

Where we stand in one screen.

36things audited

34checked live today

13working

10partly working

3broken

7stale or unused

15gaps found

What is solid

The data pipeline works end to end: county file to Airtable to sales. Airtable automations run daily without failing. The hooks that block bad commands work. The KPI weekly tab and the new site are live.

What is fragile

Nothing alerts when something breaks (3 of 9 Lambdas have an alarm). Recent work lives on one laptop. Some security items need cleanup (details kept private). The same KPI has four different answers. Deploys are by hand.

What the blueprint covers

Of 15 gaps, 6 are planned but not started, 4 are partly planned, 4 have no step and 1 conflicts with your own decision. Only 6 of 33 blueprint steps have receipts.

The map

How everything connects. Left border colour = health. Read top to bottom: data goes down.

WorkingPartlyBrokenStale / unusedNot checked

SourcesWhere data comes from

County tax rolls + appraisal portalsmanual download / open-data API

Travis clerk foreclosure PDFsold PC scraper shut down; drop-folder now

REISift (research notes)no API, browser automation

Smart Skip / Skip Geniemanual email + CSV, about a week

GHL / Pathwaize (CRM)calls, appointments, pipeline stages

↓

Build (manual, Claude sessions)Scripts people run by hand

Task K county scripts509 entry scripts, 351 with no caller

KPI + timestamp scriptsscripts/kpi, scripts/timestamps

EC2 scraper boxMcLennan / Chambers portals

Tasks A-J, L (old)no commits since Jul-Aug

↓

StoresWhere the data lives

Airtable MASTER6,035 leads, 415 fields in one table

13 supporting Airtable tablessnapshots, overrides, targets, copies

Google Sheetscounty sheets, KPI sheet, RACI sheet

S3 (task-m state)offers log, dashboard json

↓

AutomationsThings that run without a person

49 Airtable automations44 deployed, 5 off

9 AWS Lambdas + EventBridgehand-deployed, no alarms

Make.com (via "AT to Make")fires often, downstream never inspected

↓

Outputs / peopleWhat people look at

5 Airtable dashboardsmanual copy-paste to deploy

KPI Google Sheetweekly tab live, Daily tabs dead

Decision Mode site (Cloudflare)plan + feedback + chat, new

Task M dashboard (Lambda URL)nobody calls it

↓

Control planeHow the work is governed

Claude Code hooks + skills5 hook types, 13 skills, guard blocks bad commands

Receipts + ledger + capabilities6 of 33 blueprint steps have receipts

GitHub repo (private)no CI, 28 commits not pushed

Every connection, one line each

From	To	How
County files / portals	Task K scripts	manual download or API
Task K scripts	County Google Sheet	script writes tabs
County Sheet "For Smart Skip" tab	Smart Skip	manual email + CSV
Smart Skip results	Task K scripts (deceased, occupancy, tier)	manual
Task K push scripts	Airtable MASTER	REST, 25 per batch, needs sign-off
Airtable formulas + [RS-Auto] automations	Research Priority	blank-to-filled only
Airtable Automation 2	Jera / Carmen queues	Mondays 6 AM CT, tops up to 125
GHL	Airtable (daily stage sync, call KPI, dates-sync Lambda)	API pull on a schedule
GHL "Opportunity Changed" webhook	opportunity-mirror Lambda	webhook, 11 s to 7+ min late
EventBridge (2 min)	mirror-sweep Lambda	backstop for the webhook
Airtable	Make.com	"AT to Make" automation, downstream unknown
Airtable + GHL	KPI Google Sheet	scripts/kpi, run by hand weekly
Airtable dashboards .tsx files	Airtable Interface pages	manual copy-paste, no API
Claude sessions	Everything above	hooks check each command; receipts prove steps

Inventory: what we built and how it is doing

Filter by status. "Checked 27 Oct" means I looked at the running system. "From our docs" means the docs say so and I did not re-check.

Name	What it does	Status	What I found	Action
MASTER table Airtable	6,035 leads, 415 fields. The one table everything reads.	Working checked 27 Oct	Record count read today. 415 fields in one table is a risk: nobody can name what each field is for.	Keep
Stamp automations (dates entered/left) Airtable	14 small automations that write a date when a status changes.	Partly working checked 27 Oct	Two automations stamp nearly the same Deal Review date into two fields (687 vs 111 leads filled), so the same KPI gives two answers.	Needs a decision
Automation 2 (weekly top-up) Airtable	Mondays 6 AM CT, fills each researcher to 125 open leads.	Working checked 27 Oct	No failed runs in its history; 116 leads assigned at 6:00 AM on 10/26.	Watch
GHL Call KPI - Daily Airtable	Daily call counts into the GHL Call KPI table.	Partly working checked 27 Oct	Runs fine (last 3 days ok), but the numbers do not match Nakul's method: 10/23 shows 122 call messages, GHL on Yader's numbers shows 0 calls; week 10/19-25 shows 1,045 vs 472.	Fix
Sync GHL Pipeline Stages (daily) Airtable	Copies GHL stage into 8 MASTER fields.	Working checked 27 Oct	Succeeded 10/25, 10/26, 10/27. One failure on 20 Aug.	Keep
Log Research Priority Overrides Airtable	Logs when a researcher changes a priority.	Working checked 27 Oct	Docs say broken since 4 Sep. Live runs: failed on 21 Oct, succeeding every run since 26 Oct. The docs are out of date.	Watch
Weekly Tier / Deal Review / Capacity snapshots Airtable	Mondays: freeze counts into snapshot tables.	Broken checked 27 Oct	Deal Review snapshot is cut off at 1,000 (weeks 9/21 and 10/12 identical: 1,000 unset, 1,426 total). Capacity shows Deal Review actual = 0 (date filled for 0.4% of leads).	Fix
5 undeployed automations Airtable	AUTO DATE, Deletion for Pass-off, Set Push to Sales CRM, Model D on update, hourly Sweep.	Stale checked 27 Oct	Built, never switched on. Nobody has decided: deploy or archive.	Needs a decision
"AT to Make" Airtable	Sends a record to Make.com.	Not checked checked 27 Oct	Fires many times a day (3 runs between 22:12 and 23:27 on 10/26). Nobody has inspected what Make does with it.	Needs a decision
Snapshot tables + overrides + targets Airtable	Weekly Tier, Deal Review, Capacity, Outreach, Meaningful by RM, Pipeline Aging, Stage Targets, Weekly Goals.	Partly working checked 27 Oct	Stage Targets are auto-seeded placeholders, so most dashboard health cards read a meaningless 150%.	Fix
MASTER copy table (346 fields) Airtable	A second copy of the lead table.	Stale checked 27 Oct	None of the docs I read mention it.	Needs a decision
5 Interface dashboards (.tsx) Airtable	Executive, Department, Conversion, Researcher, Pipeline Aging.	Not checked from our docs	Deploy is a manual paste. Whether the latest version is pasted is unknown. Exec Dashboard was a draft on 15 Sep.	Needs a decision
KPI sheet: weekly tab Google	Weekly KPI numbers per week, column per week.	Working checked 27 Oct	Updated by scripts; 10/19 and 10/12 checked against sources. Rows move when people insert rows, so scripts find rows by name.	Keep
KPI sheet: Daily Input tabs Google	Two tabs where people typed daily numbers.	Stale checked 27 Oct	Columns stop at Sep 30. Last typed cell: Jul 31 and Jul 28.	Retire (archive)
KPI sheet: old tabs Google	Copy of Sale Team KPi, Team Wide KPIs - Archive, Department, KPIs, and more (14+ tabs).	Stale checked 27 Oct	Several are copies or archives. Nobody has listed which are still read.	Needs a decision
County sheets (12 tabs each) Google	Per-county staging before Airtable.	Working from our docs	No write guard on Sheets.	Keep
9 Lambdas (not 10) AWS	dates-sync, meaningful-conversations, daily-checklist, pipeline-aging, mirror-sweep, webhook-test-receiver, task-m puller, task-m dashboard, task-n.	Partly working checked 27 Oct	The docs say 10 and name a separate opportunity-mirror. Only 9 exist. 0 throttles. Errors in 14 days: mirror-sweep 14 (hit its 60 s timeout on 14 Oct), meaningful-conversations 3. Sales-kpi peaked at 277 s of its 300 s limit (92%).	Watch
Schedules (7 triggers) AWS	EventBridge rules and Schedulers.	Working checked 27 Oct	All 7 enabled, none point at a missing function. 5 already use America/Chicago time. Only the daily-checklist (cron in UTC) will shift an hour on 1 Nov.	Fix
Error alarms AWS	CloudWatch alarms with an email topic.	Partly working checked 27 Oct	3 alarms exist (dates-sync, meaningful-conversations, pipeline-aging), all OK. The docs said none. The other 6 Lambdas have no alarm, including mirror-sweep (14 errors) and sales-kpi (92% of its time limit).	Fix
task-m dashboard Lambda AWS	HTML KPI page.	Unused checked 27 Oct	0 calls in 14 days. Candidate to retire.	Retire (archive)
webhook-test-receiver AWS	Named a test stub.	Partly working checked 27 Oct	In real use (229 calls in 14 days) despite the test name. Access rules need review (details private).	Needs a decision
EC2 montgomery-scraper (t3.small) AWS	County portal scraping.	Working checked 27 Oct	Running, launched 19 Oct. Access rules need tightening (details private).	Fix
EC2 hermes-agent (t3.micro) AWS	Unknown. In no doc.	Not checked checked 27 Oct	Running since 9 Aug. Nobody has documented what it does. Access rules need tightening (details private).	Needs a decision
AWS access keys AWS	Account access keys.	Broken checked 27 Oct	Need cleanup. Details kept private for security.	Fix
Secrets Manager (8), S3 (1), logs (9) AWS	Credentials, one 26 MB bucket, Lambda logs.	Working checked 27 Oct	Secrets cost about $3 a month. Logs never expire but total 20 MB, so negligible.	Keep
src/task-k county scripts Repo	All county work. 822 files.	Partly working checked 27 Oct	Audit on 4 Oct: 351 of 509 entry scripts have no caller and 0 tests. Per-county copies of the same rules (government-owner filter, groups).	Fix
Tasks A-J, L (old) Repo	First-generation tasks.	Unused checked 27 Oct	Last commit Jul-Aug 2026. Some unfinished (Task A Jr RM, Task E recovery).	Needs a decision
Loose files at the repo root Repo	19 _scratch files, build scripts, json.	Stale checked 27 Oct	All tracked in git.	Retire (archive)
output/ folder Repo	2,661 files, 346 MB (task-k 286 MB).	Partly working checked 27 Oct	Mix of results and throwaways, not separated.	Needs a decision
git on GitHub (private) Repo	The backup and history.	Partly working checked 27 Oct	28 commits (24-27 Oct) are only on this Mac. No CI, no branch protection (GitHub Free).	Fix
Secrets scan report Repo	Report of a secrets scan.	Broken checked 27 Oct	Needs cleanup. Details kept private for security.	Needs a decision
Hooks (5 types, 7 files) Claude control plane	Block bad commands, save context, check on stop.	Working checked 27 Oct	Blocked a deploy flag and a delete command in this very session. They work.	Keep
Skills (13) + knowledge/ + STATE/GATES Claude control plane	Memory and routines between sessions.	Working checked 27 Oct	Some docs older than code (INDEX, ARCHITECTURE, systems pages).	Fix
Receipts + capabilities.json Claude control plane	Proof a blueprint step is done.	Partly working checked 27 Oct	6 receipts (P0.0, P0.1, P0.2, P0.3, P1.1, P2.5). 27 capabilities, only 3 required.	Keep
Decision Mode site (Cloudflare) New this week	Lead Speed plan, feedback on every section, chat.	Working checked 27 Oct	Free tier. Name-only sign-in. Chat is keyword lookup; AI binding not added.	Keep
Lead-speed stage log prototype New this week	Pure code, 17 tests.	Working checked 27 Oct	Runs locally, writes nothing.	Keep

How the automations perform

Numbers from today, not from the docs.

Airtable (49 automations)

44 deployed, 5 switched off.
Automation 2: no failed runs in history. It assigned 116 leads at 6:00 AM on 10/26.
GHL Call KPI (daily): ok 24, 25 and 26 Oct. Last failure 19 Aug (secret not found).
Sync GHL Pipeline Stages (daily): ok 25, 26, 27 Oct. Last failure 20 Aug.
Log Research Priority Overrides: failed 21 Oct, ok on every run since 26 Oct.
"AT to Make": runs about every 35 minutes on busy days.

Do the numbers they produce agree with the source?

No. Weekly Deal Review snapshot stops at 1,000 leads, so two different weeks show the same total (1,426).
No. The GHL Call KPI table counts 1,045 for the week of 10/19. GHL on Yader's two numbers shows 472 calls.
Partly. Capacity snapshot: Deal Review actual is 0 because the date it reads is filled for 0.4% of leads.

AWS

Checked live in AWS (read-only), Tue 27 Oct 2026, read-only

9Lambdas (docs said 10)

7schedules, all enabled

3 of 9Lambdas with an alarm

2EC2 servers

$27.80gross cost, 30 days

$0net bill (credits)

Lambdas, last 14 days

Function	Calls	Errors	Slowest run	Limit	Note
dates-sync	14	0	30 s	60 s	ok, daily 9:00 Chicago
meaningful-conversations	16	3	90 s	300 s	3 errors on 16 Oct, cause not found; daily 9:30 Chicago (docs said weekly)
daily-checklist	15	0	240 s	300 s	UTC cron, shifts on 1 Nov; no alarm
pipeline-aging-refresh	14	0	9 s	90 s	ok, daily 9:15 Chicago
task-m sales-kpi	14	0	277 s	300 s	92% of limit, no alarm
task-n outreach	2	0	623 s	850 s	weekly Monday, as designed
opportunity-mirror-sweep	10,292	14	60 s (timeout)	60 s	every 2 min; 14 errors on 14 Oct; no alarm
webhook-test-receiver	229	0	4 s	30 s	real traffic, named test
task-m dashboard	0	0	-	15 s	unused

Cost, last 30 days (gross usage)

Service	USD
EC2 compute	15.92
VPC	5.62
Secrets Manager (8 secrets)	3.08
EBS and other EC2	3.05
Cost Explorer	0.14
Lambda, EventBridge, S3, CloudWatch	about 0
Credits applied	-27.80
Net charged	0.00

Docs were wrong in 3 places. Docs say 10 Lambdas (9 exist), say no alarms (3 exist) and say no schedule uses Chicago time (5 of 7 do).

Credit cliff. The real run rate is about $28 a month. It becomes a bill when credits end. Expiry not checked.

Stale, unused or needing a decision

Nothing here is deleted. "Retire" means archive. Tell me keep or retire for each.

Name	Why	Suggested
Stamp automations (dates entered/left)	Two automations stamp nearly the same Deal Review date into two fields (687 vs 111 leads filled), so the same KPI gives two answers.	Needs a decision
5 undeployed automations	Built, never switched on. Nobody has decided: deploy or archive.	Needs a decision
"AT to Make"	Fires many times a day (3 runs between 22:12 and 23:27 on 10/26). Nobody has inspected what Make does with it.	Needs a decision
MASTER copy table (346 fields)	None of the docs I read mention it.	Needs a decision
5 Interface dashboards (.tsx)	Deploy is a manual paste. Whether the latest version is pasted is unknown. Exec Dashboard was a draft on 15 Sep.	Needs a decision
KPI sheet: Daily Input tabs	Columns stop at Sep 30. Last typed cell: Jul 31 and Jul 28.	Retire (archive)
KPI sheet: old tabs	Several are copies or archives. Nobody has listed which are still read.	Needs a decision
task-m dashboard Lambda	0 calls in 14 days. Candidate to retire.	Retire (archive)
webhook-test-receiver	In real use (229 calls in 14 days) despite the test name. Access rules need review (details private).	Needs a decision
EC2 hermes-agent (t3.micro)	Running since 9 Aug. Nobody has documented what it does. Access rules need tightening (details private).	Needs a decision
Tasks A-J, L (old)	Last commit Jul-Aug 2026. Some unfinished (Task A Jr RM, Task E recovery).	Needs a decision
Loose files at the repo root	All tracked in git.	Retire (archive)
output/ folder	Mix of results and throwaways, not separated.	Needs a decision
Secrets scan report	Needs cleanup. Details kept private for security.	Needs a decision

How we work, and what we use

The working method, then the resources.

1. You decide

Harsh is the only approver. Ralph is the SOP and review authority. Nakul owns KPI definitions.

2. One Claude session, one task

Each task runs in its own session. Memory carries over through knowledge/STATE.md, GATES.md and a ledger.

3. Hooks stand guard

Every command is checked. Deletes, live writes and secrets are blocked or need a plan id.

4. Scripts do the work

Read-only first, dry run, backup, write, read back. Rows are found by name, never by number.

5. Proof before "done"

A receipt must pass. Without one the step is not done, however green the tests look.

6. Humans handle the gaps

Smart Skip emails, dashboard pastes, EOD typing, merges.

Resources

Resource	Used for	Cost known?
Smart Skip / Skip Genie	Skip tracing	About $0.16 a row (Montgomery Group 1 about $258)
AWS (acct ending 2485)	10 Lambdas, EventBridge, S3, one EC2	$27.80 a month gross (EC2 $15.92, VPC $5.62, Secrets $3.08, EBS $3.05). Credits cover all of it; net bill $0. Credit expiry not checked.
Airtable	Lead table, automations, dashboards	Plan and bill not stated
GHL / Pathwaize	Sales CRM	Plan and bill not stated
Cloudflare Pages	This site, 2 key-value stores	Free tier
GitHub (Free)	Backup and history	Free. Branch protection needs Pro, about $4 a month
Google Workspace	Sheets, Docs, service account	Not stated
NVIDIA vision API	Occupancy check from Street View	Not stated
CapSolver	Travis court portal captcha; key needs rotation	Not stated
Make.com	Receives "AT to Make"	Account and plan unknown
Monday.com	Task F source; connector disabled by admin	Unused
RentCast (free tier)	Phase 2 foreclosure, waiting on a key	Free
Claude subscription	All sessions	Not stated

Gaps, and is there a fix in the blueprint?

Blue = the blueprint plans it but has not started. Red = no fix planned or it conflicts with a decision.

G1. Nothing watches the automations

Evidence
Only 3 of 9 Lambdas have an error alarm. mirror-sweep (14 errors, hit its timeout) and sales-kpi (92% of its time limit) have none. No Airtable automation has any alert; the overrides log was failing from 4 Sep until the 20 Oct fix.

Blueprint
P6.5 daily check + dead-man alarms; P6.2 scheduler Planned
Not started. The 24x7 layer does not exist (no ops/ folder).

What I recommend
Build a small read-only morning check first (it reads Lambda errors, Airtable runs, sheet freshness). Do not wait for the full P6 build.

G2. One laptop holds recent work

Evidence
28 commits (24-27 Oct) exist only on this Mac. The KPI fixes, plan, prototype and this audit are in them.

Blueprint
P0.2 safety floor, P2.5 new-laptop bootstrap Partly planned
Backup push is not a blueprint step.

What I recommend
Push to a claude/ branch today (needs your Y).

G3. Secrets and keys need cleanup

Evidence
Details kept private for security. Owner: Harsh.

Blueprint
P8 security hardening Conflicts with a decision
Blueprint and your decision contradict each other.

What I recommend
Reviewed with Harsh privately.

G4. The same KPI has four answers

Evidence
GHL Call KPI table, Task N Lambda, Meaningful-by-RM table and the sheet all count calls differently (week 10/19: 1,045 vs 472 vs 378 vs 462). Week meaning differs too (current week vs previous week).

Blueprint
P5.1 KPI registry, P3.1 reconcile Planned
Not started. Week-meaning mismatch is not in the blueprint.

What I recommend
Our KPI daily/weekly plan is the practical start: one definition per KPI, one source, a morning recount.

G5. Lead-speed dates are missing or overwritten

Evidence
Only 2 stage ends are trustworthy. Date Entered Research is 88% bulk-stamped. Assignment date is overwritten.

Blueprint
Not covered (STATE mentions a "Phase Time Log" as the fix) No fix planned
No blueprint step.

What I recommend
Our Lead Speed plan fills this. Add it to the blueprint as its own step.

G6. Business rules are copied in 4+ places

Evidence
The P0-P14 tier waterfall lives in code, an Airtable formula, two Lambdas and a dashboard file. Government-owner filter is copied per county. A snapshot file the sync scripts need is missing, so two scripts crash on this Mac.

Blueprint
P5.2 county framework (partly) Partly planned
Tier duplication has no step.

What I recommend
Pick one source of truth for tiers and add a test that fails when the copies disagree.

G7. Scripts bypass the write guard

Evidence
129 Task K scripts call Airtable directly. The guard only blocks the old bases, not the live one.

Blueprint
P4.1 write gateway Planned
Not started.

What I recommend
Keep the hook guard as the interim control. Gateway when the daily check exists.

G8. Daylight saving on 1 Nov

Evidence
Of 7 triggers, 5 already use America/Chicago. Only the daily-checklist rule (cron in UTC) will run an hour earlier after 1 Nov (6 AM becomes 5 AM). Airtable automations also need a check on Mon 2 Nov. 5 days away.

Blueprint
P6.2 EventBridge Scheduler with America/Chicago Planned
Not started.

What I recommend
Move the checklist to a Scheduler entry with America/Chicago, like the other five. Check Automation 2 and the snapshots on Mon 2 Nov.

G9. Dead and half-built things are not retired

Evidence
5 undeployed automations, Tasks A-J unused, Daily Input tabs dead, 19 scratch files, a public unused Lambda URL, an unexplained MASTER copy.

Blueprint
P0.3 lists them; no retire step No fix planned
No archive pass is planned.

What I recommend
One archive pass: move, never delete. Needs a keep/retire decision per row on this page.

G10. Deploys are by hand, drift goes unseen

Evidence
No IaC, no CI. Two Lambdas ran old tier code until redeployed on 22 Oct. Docs list 10 Lambdas, AWS has 9.

Blueprint
P1.5 CI (blocked by GitHub Free), P8 least privilege Partly planned
Drift check is not a step.

What I recommend
Add a deployed-code-equals-repo check to the morning check.

G11. Docs disagree with each other and with the system

Evidence
Overrides log "broken" in docs, working live. INDEX says meaningful-conversations broken, its page says fixed. ARCHITECTURE says 4 capabilities, there are 27. Tier range P0-P11 vs P0-P14.

Blueprint
P2.1 knowledge consolidation Planned
Partial.

What I recommend
Make this page the status source and date every claim.

G12. Cost is invisible

Evidence
AWS usage is $27.80 a month gross (EC2 $15.92, VPC $5.62, Secrets $3.08, EBS $3.05). Credits cover all of it, so the bill is $0 today. Nobody has checked when the credits run out. Airtable, GHL and Google costs are not stated anywhere.

Blueprint
Not covered No fix planned
No step.

What I recommend
Find the credit balance and expiry in the AWS console. Add Airtable and GHL plan costs.

G13. Human steps are hidden single points

Evidence
Dashboard deploy (paste), Smart Skip (email), EOD numbers (typed), GitHub merges (you). Each waits on one person.

Blueprint
P6.6 dashboard reader (partly) Partly planned
Smart Skip has no API, so it stays manual.

What I recommend
Name an owner and a backup for each, on the board.

G14. AWS access needs tightening

Evidence
Details kept private for security.

Blueprint
P8 security hardening (least privilege, IAM) Planned
Not started.

What I recommend
Reviewed with Harsh privately.

G15. A running server nobody documented, and a credit cliff

Evidence
EC2 "hermes-agent" has run since 9 Aug and appears in no doc. Credits currently pay the whole $27.80 a month; expiry unknown.

Blueprint
Not covered No fix planned
No step.

What I recommend
Tell me what hermes-agent is. If it is unused, stop it (not delete). Check the credit expiry date.

Blueprint scoreboard

The blueprint has about 33 steps. Receipts exist for 6. They were run on 22-23 Oct and pass today only if re-run inside 7 days.

Done, with receipt (6)

P0.0 Referee
P0.1 Dead-feature audit
P0.2 Safety floor
P0.3 Seed capabilities
P1.1 Settings + hooks
P2.5 New laptop

Built, no receipt (3)

P1.2 Receipts self-test
P2.2 Compaction-proof
P2.3 Skills

Blocked (2)

P1.5 CI judge (GitHub Free returns 403 on branch protection)
P1.4 Seal (only you can do it)

Not started

P3 Verify suite, ledger, registry
P4 Write gateway, Jev/Laya
P5 KPI registry, county framework
P6 Agents, daily check, TODAY page
P7 Scale
P8 Security hardening

What to do first

In order. The first three take under an hour each.

Push the 28 unpushed commits to a backup branch (needs your Y). Removes the one-laptop risk.
Security cleanup (details are shared privately with Harsh).
Before 1 Nov: fix the checklist schedule to Chicago time, and check Automation 2 on Mon 2 Nov.
Build the read-only morning check: Lambda errors, Airtable runs, sheet freshness, one message a day. This closes G1 without the full blueprint.
Fix the numbers: repair the 1,000 cap in the snapshot, pick one Deal Review date, then the KPI daily tab.
One archive pass on everything marked retire or decide.
Add Lead Speed and this audit to the blueprint as steps, so the plan covers what we actually do.