The adversarial review that attacked the plan — 22 findings.
25 — Red-Team Review
Prepared 2026-07-22. Reviewer mandate: MASTER_PROMPT.md "Red-team requirement" (lines ~1027–1052). Method: full adversarial read of files 00–24, 27, research_notes/DECISION_BRIEF.md, the data/ artifacts, and the decision_gate/ directory, followed by independent recomputation of the funnel math (file 22 vs files 11/23), a compliance sweep of every script and calendar block against the plan's own guardrails (file 05), and a cross-file consistency check of dates, targets, and claims. I attacked the plan; I did not invent problems. Where the plan genuinely holds up, that is stated at the end.
Severity key: CRITICAL = would cause plan failure or compliance exposure if executed as written. MAJOR = materially weakens outcomes. MINOR = polish.
CRITICAL FINDINGS
C1. Required deliverables are missing — the project fails its own completion standard
What:START_HERE.md does not exist. 26_revised_final_plan.md does not exist. decision_gate/ is empty — all three required files (decisions_needed_before_build.md, compliance_approval_checklist.md, prompt_2_readiness.md) are absent. File 27 even references "the compliance-approval checklist in decision_gate/" (P0-5) — a pointer to a file that was never written.
Where: MASTER_PROMPT.md required-output tree (lines 948–990) and completion standard (lines 1060–1076: "The project has a clear decision gate for Prompt 2"); ls decision_gate/ returns nothing.
Why it matters: Prompt 2 is explicitly instructed to read 26_revised_final_plan.md and the two decision-gate files before building. As delivered, Prompt 2 has no input, the "5-minute orientation" document Mitchell is supposed to start with doesn't exist, and the completion standard is unmet on at least three counts. The master prompt also requires that "the final recommendation must explain what changed after red-team review" — impossible without file 26.
Fix: In the revision pass, write all four missing files. START_HERE.md per the MASTER_PROMPT spec (one-sentence strategy, three don'ts, first ten actions, week-1 calendar, first five metrics, five unresolved compliance questions, links). 26 must include a "what changed after red-team" section addressing this review by finding number.
C2. The financial model contradicts itself: the hand-set loan ramps run 2–3× above what the model's own funnel inputs can produce
What: File 22's base-case funnel inputs, multiplied through, produce roughly 5 funded loans in year 1 — not 14. Base inputs: 4 scenario requests/mo (reached "by month 3") × 35% scenario→borrower × 70% borrower→app ≈ 1.0 agent-engine app/mo; consumer engine 4 leads/mo × 30% ≈ 1.2 apps/mo from month 4. Year-1 apps ≈ 20–21. At the model's own derived 27% pull-through, that is ≈ 5–6 funded loans with the 1.5-month lag — which is what the file labels the conservative ramp (6). The "base" ramp (14) implies ~50 applications, ~2.4× the base funnel's capacity; a true conservative case (conservative inputs: 2 scenarios/mo, 25%, 60%, 22% pull-through) is 1–3 loans, i.e., a year-1 income that rounds to zero.
Where: 22_financial_model_assumptions.md §1–§3 (inputs) vs §6 (ramps). The file even says the workbook "flags it" when hand-set ramps exceed funnel capacity — the flag condition is met inside the file itself and was not caught.
Why it matters: Every downstream expectation — the 30/60/90 gates, the risk register's "survivable if expected" framing, and any household budgeting Mitchell does — is calibrated to numbers the model's own machinery says are optimistic-to-aggressive. The user-facing story ("6 conservative / 14 base") materially overstates the plan's self-consistent expectation. For a reserved newcomer with no book, the honest headline is: base ≈ 5, conservative ≈ 1–3, and months 1–5 ≈ $0.
Fix: Relabel the ramps: current "conservative" (6) becomes base; current "base" (14) becomes aggressive; build a true conservative ramp from conservative inputs (1–3). Recalibrate the file-11 gates and file-23 color bands to the corrected base. State the corrected numbers in files 01, 22, and the workbook.
C3. No personal-runway analysis exists anywhere — the plan's most likely kill condition is unexamined
What: Risk R11 names the ramp-income gap and even contains the placeholder [RUNWAY_MONTHS] — but no file computes, estimates, or even asks for Mitchell's monthly personal burn rate, savings, household support, or health-insurance situation. File 27 (the questions-for-Mitchell file, whose entire purpose is questions only humans can answer) has no runway question. File 20 asks "draw vs. commission-only" (Q14) but never asks the draw amount, benefits/health insurance, or W-2 vs 1099 status (which file 22 separately flags as unknown via [TAX_PCT]).
Why it matters: Per C2, the honest expectation is near-zero income for two-plus quarters. If Mitchell has 3 months of runway, this plan is not viable as designed regardless of its quality, and the correct plan includes an income bridge — a question the plan never surfaces. R11's own early-warning signal ("runway below [RUNWAY_MONTHS]") can never fire because the variable is never filled. Financial fragility is a named red-team dimension in the master prompt; the plan gestures at it and then doesn't do the arithmetic.
Fix: Add P0-0 to file 27: monthly personal burn, liquid runway in months, health-insurance plan, and the go/no-go rule ("if runway < 9 months at conservative-case income, define the bridge before Day 1 — draw negotiation, part-time income floor, or household support in writing"). Add draw amount and benefits to file 20 Block 4. Fill [RUNWAY_MONTHS] in R11 and add a runway line to the workbook.
C4. The scripts assert "reviewed with my team" and a Saturday SLA as facts while the reviewer and weekend coverage are [UNVERIFIED] — a self-made misrepresentation trap
What: Script #6 states, in the voice Mitchell will use on live calls: "I review every brief with my team at Ready before it goes out — you get a checked answer, not a guess." Objection O2 repeats it ("every answer I give is reviewed with my experienced team at Ready"). Meanwhile file 09 §2.3 marks reviewer availability [UNVERIFIED — confirm reviewer availability with manager] with a 60-minute review-turnaround budget, and file 20 Q31/Q36 treat "can Ready back the 4-hour promise" as an open question. The published SLA also covers Saturday 9:00–1:00 (09 §2.1, §2.5) — weekend ops coverage is unverified (file 20 Q15) and a ~13–30-person Miami lender plausibly has none.
Why it matters: The plan's own compliance file (05 §A.5) notes the 2024 TSR amendments removed the B2B exemption from misrepresentation rules. If the review gate is Mitchell reading his own draft twice because no senior LO is staffed to it — or if Saturday briefs go out unreviewed — the "human-reviewed by my team" claim is a misrepresentation made on a solicitation call, and the SLA becomes the "broken promise worse than no promise" that B4 warns about. This is the exact class of claim the plan spends pages banning ("no speed/approval claims without operational proof") — and it survived into the flagship scripts.
Fix: Two script versions: v0 (pre-verification) says "I check every brief against the guidelines and my team before I'd act on it" with no review-gate claim; v1 (post-verification, named reviewer confirmed in writing) may say "reviewed with my team." Drop Saturday from the published SLA until weekend coverage is confirmed; Saturday requests ack same-day, brief next business morning (mode-2 already contains this language — make it the published standard). Add "review-gate claim verified?" to the pre-outreach checklist in file 10.
C5. The entire first 30 days hangs on a small, demonstrably inattentive employer approving everything within 72 hours — with no contingency path
What: File 10 requires: manager meeting booked Day 1 at 10:00 AM; written IL confirmation and product sheets within 48 hours; compliance approval of call script, email, signature, bio, and landing copy by Hour 72 — from a ~13–30-person Florida lender whose public evidence of marketing/compliance bandwidth is a license page stale for 8+ months and lorem-ipsum placeholders on its partners page (facts the plan itself established, 03 claims #3/#7). The plan's hard rule — correctly — is "no calling on unapproved copy." But the only contingency offered is "chase politely and substitute drills," and files 11–12 present "Week 2: all 30 agents in motion" as the plan of record.
Where: 10_first_72_hours.md items 2, 4, 10, 11 and the "hard rule"; 12_four_week_ramp_calendar.md Weeks 1–2; 03 §3, §7; R15 (which names the bottleneck but scopes it to content only).
Why it matters: The single most probable first-30-days outcome is approval starvation: the meeting slips, the "written IL product menu" comes back as shrugs and "let me check," the disclosure block takes two weeks, and the ramp calendar silently becomes fiction while Mitchell drills alone. There is a real possibility Ready has never produced an IL-compliant disclosure block or reviewed an IL ad in its life (Mitchell may be IL employee #1 — P1-5). R15's mitigation ("agree the approval path up front") is right but is itself gated on the same meeting.
Fix: Add an explicit approval-starved contingency track to files 10–12: (a) Day-1 fallback if the meeting can't book inside 72 hours (video call with whoever owns compliance, even 30 minutes, on P0-1 only); (b) a defined "week 1–N holding pattern" (drills, list-building, sample-brief portfolio, facts library, workshop scouting — all no-approval-needed work) with a named escalation at day 7 and day 14; (c) a trigger: if no approved outreach copy exists by day 21, that is a strategy-level employer-viability signal feeding the C6 hedge, not just a delay.
MAJOR FINDINGS
M1. No employer hedge: Ready may be the weak link, and the plan has no exit criteria, no alternate-sponsor research, and no entry for the December 2026 IL license renewal
What: The plan documents Ready's fragility thoroughly (R5–R8: thin ops, stale site, unverified menu, no IHDA) and file 09 §6 concedes "if Ready cannot support the wedge's product needs by ~day 90, that is a strategy-level conversation" — and then never has the conversation. There is no walk-away trigger list, no survey of alternative IL sponsors who hire new MLOs (the research already names the natural candidates: the IHDA-participating lenders in 07's consequence analysis), and no plan for portability of the assets (domain ownership, contact export, and content ownership are single meeting questions — 20 Q16/Q19 — with no downstream planning). Worse: Ready's IL license expires 2026-12-31 (A1), i.e., in month ~5 of the plan. Non-renewal — entirely plausible for a lender that has apparently done nothing with the license in 8 months — would kill the book mid-ramp. The risk register has no entry for it; A1 buries it as "re-check at renewal."
Why it matters: The plan treats Ready problems as questions to ask, not scenarios to plan. If the manager meeting returns "FHA not delegated in IL / no DPA pairing / no reviewer / no marketing support," the plan's product spine, SLA, and content engine all fail simultaneously — and Mitchell, a licensed MLO with a growing local asset base, has zero prepared moves.
Fix: Add to file 24: R20 (IL license non-renewal, review 2026-11-01) and R21 (employer cannot support the plan). Write a half-page "Plan B" section in file 26: three named walk-away triggers (e.g., no written IL product menu by day 30; no IHDA/DPA answer + no FHA delegation by day 60; SLA back end unstaffable by day 30), a pre-researched shortlist of 3–4 alternative IL sponsors, and a checklist of what must be secured now to make a future move survivable (personal domain, personally-owned CRM export rights in writing, content copyright position, personal Google Business Profile).
M2. Cross-file target incoherence: the 30/60/90 gates, the KPI scorecard, and the financial model describe three different businesses
What: File 11 day-30 targets: 8 scenario requests, 20 meaningful conversations, 2 apps; day-60: 20 scenarios, 6 apps; day-90: 12 apps, 2–3 closings. File 22 base: 4 scenario requests/mo reached by month 3, ~1 agent app/mo, and MONTH_FIRST_FUNDED = 4 with the explicit sentence "a first funded loan before month 3 would be luck, not plan" — while file 11's day-90 gate demands 2 closings (month 3) as the scale criterion. File 23 L5 sets scenario-request green at 1/wk from month 3+ — file 11 demands ~2/wk in month 1. D2's stop rule uses <6 scenarios at day 45; file 11's gate uses <6 at day 30.
Why it matters: Whichever numbers Mitchell believes, the other files call him wrong. At day 30 a performance the financial model would call on-plan (2–4 scenarios, ~12 conversations) reads as a double-red failure under file 11, triggering pilot-list surgery and script rewrites that aren't warranted — or teaching him to ignore gates entirely, which destroys the plan's best feature (pre-instrumented stop rules). The day-90 "≥2 closings → scale" gate is calibrated to luck by the model's own definition.
Fix: One source of truth. Rebase all file-11 targets on corrected file-22 base (per C2): day-30 ≈ 10–12 meaningful conversations, 3–4 scenario requests, 1 app; day-60 ≈ 8–10 scenarios cumulative, 3 apps, 1–2 preapprovals; day-90 gate on leading indicators (A-tier agents, repeat scenario senders, SLA record) with closings explicitly excluded from gating before month 4. Align 23's L5 and D2's stop-rule dates to the same numbers.
M3. The acquisition engine is cold outreach in a Scenario Desk costume — and the plan's one genuine warm network is treated as a conflict instead of an asset
What: Strip the branding and month 1 is a 14-day cold sequence (email → call → voicemail → LinkedIn → asset → drop-in) run on 30 strangers — the generic playbook every sales trainer sells, executed at ~50 dials/week (12_ weeks 2–4) by a reserved newcomer. The differentiation (the brief) only exists after an agent bites; the sequence gives agent #14 no proof, just a promise. The plan's own answer — "can I send you one example brief?" — is buried as a second-fallback line in §2.6 rather than being the lead mechanic. Meanwhile, the family connection to Crosstown (~150–160 agents, Tinley HQ) — the one warm network a no-contacts introvert actually has — is allocated a share of just 4 "independent" slots and is discussed mostly as a Flex-conflict flag (15 §1). The Flex caution is correct; extending it into near-total non-use of Hybrid-tier and non-Flex Crosstown agents, plus family introductions to agents at other brokerages, is strategy malpractice for this specific person. Scorecard 08 rates Model D "sustainability 9" for the reserved personality — a score that describes the desk's steady state while hiding that its months 1–2 are Model A cold networking with better scripts.
Where: 15 §1 (slot allocation, Crosstown flag), §4; 16 §2.6; 12 weeks 2–4 dial counts; 08 Model D scoring.
Why it matters: Cold email + cold call response rates for busy realtors are low single digits; a reserved rookie's will be lower. The most probable month-1 outcome is 30 completed sequences, a handful of polite conversations, 0–2 scenarios — which under M2's miscalibrated gates reads as failure, and psychologically lands as "the plan doesn't work," in week 4, on the person least suited to grinding cold calls.
Fix: (a) Lead every first touch with a finished sample brief (2–3 polished synthetic briefs exist by Day 4 per file 10 item 8 — attach one; the product demos itself). (b) Re-slot the pilot: 6–8 warm entries first — Crosstown Hybrid/non-Flex agents via family intro, plus family-network introductions to agents elsewhere — cold corridor second. Warm-first is both higher-yield and personality-correct. (c) Rescore 08's sustainability row honestly (D = 7 with a note that its ramp phase is high-social) so the tradeoff is visible.
M4. The DNC-scrub dependency can silently kill the calling channel in week 1
What: The plan's (correct) conservative posture: treat realtor cells as DNC-protected, scrub everything that isn't a verified office landline (05 §A.3, B7). Scrubbing requires an organizational National DNC Registry Subscription Account Number — which Ready, a small FL lender with no visible outbound program, quite plausibly does not hold ([UNVERIFIED], P1-3, 20 Q18). If it doesn't: under the plan's own rules Mitchell may call verified office landlines only — mostly receptionists — while the calendar has him dialing 10 agents from Day 3 and file 10 item 9 requires "every row DNC-scrubbed" by Hour 48, an impossibility without the SAN.
Why it matters: The 14-day sequence loses its highest-conversion touch (live call) exactly when the plan needs early wins, and nobody has priced that scenario: the connect-rate assumptions behind "5 meaningful conversations/week" implicitly assume cell access.
Fix: Promote P1-3 to P0 (it blocks Day-3 calling, not week 2). Add the fallback to files 12/15 explicitly: until a SAN exists, the sequence runs email/LinkedIn/office-landline/in-person, with expectations cut accordingly; Ready obtaining a SAN (first 5 area codes are low-cost) becomes a Day-1 written request with an owner and date on the Decision Sheet.
M5. Compliance leaks that survived into the scripts and calendars
Four items a CCO would strike or condition, all violations of the plan's own rules:
Script #7 offers a listing-specific one-pager. "If you ever want a financing one-pager for a listing… about what buyers of this kind of home should know — I'm glad to make one you can point people to." File 05 §F trap 2 draws the line at whose expense is defrayed and whose product is promoted — a flyer keyed to the agent's listing markets the listing and substitutes for material the agent would otherwise produce. It is materially different from the approved "my own branded flyer about my own services." (16 script #7 vs 05 §F.2.) Fix: restrict the offer to Mitchell's generic financing one-pagers (condo checklist, DPA sheet, payment anatomy); kill "for a listing / this kind of home" phrasing; send the revised script for written CCO sign-off.
The forwardable summary is designed to strip disclosures. 09 §2.3 step 4: brief delivered as PDF + "three-sentence plain-text summary in the message body (agents forward the summary to clients; the PDF carries the disclosures)." A summary that names product terms ("FHA at 3.5% down clears this" — 09 §2.4 example; a down-payment percentage is a Reg Z trigger term) is architected to reach a consumer detached from its disclosure block. (09 §2.3–2.4 vs 05 §G.1/G.7.) Fix: the plain-text summary must be terms-free (conclusion + "full brief attached with details and disclosures"), or carry the abbreviated block itself; state the rule in the brief SOP.
Open-house "payment math at ask price." 12 (week-3 Saturday) and 14 (field prep) direct Mitchell to be useful with "payment math at 6.55% on ask price" — live rate-and-payment quoting to consumers, by a rookie, on unapproved assumptions, one step from being written down; the plan elsewhere bans self-produced rate/payment content outright and gates all payment illustrations behind F3's approval machinery (18 F3). Fix: field talk track = process, DPA structure, tax anatomy, and "I'll send the compliant worked example" — no live numbers until the approved-assumptions block exists.
Hardcoded program facts in a live call script. O6 quotes "$25,000… reopened July 20th" as evergreen speech while A14's own rule is "funding cycles drain — confirm live before every borrower conversation." (16 O6 vs 00 A14.) Fix: bracket the figures [CONFIRM LIVE] in the script the way every other file brackets unknowns.
M6. The condo-triage "publish first, beat August 3" plan is arithmetically dead on arrival — and file 09 doesn't know what file 11 knows
What: The Aug-3 Limited-Review sunset is 12 days after the prep date. The asset requires: Lab naming approval (or fallback), the disclosure block (P0-1), compliance review with an unknown turnaround (R15), and 4–6 hours of writing — behind a Day-1 meeting that may itself slip (C5). File 11 Phase 2 quietly concedes the reframe ("the sunset hit 2026-08-03… reframe live"); file 09 §3.3 and E1 still command "condo triage (wk 1–2, beats the 08-03 deadline)" and 06 §5 sells the pre-deadline urgency.
Why it matters: The strongest of the three "timely hooks" will almost certainly ship after its deadline; the exec-level story (three urgent 2026 hooks) quietly degrades to one-and-a-half. Not fatal — "here's what just changed and what to do now" is still good content — but the plan should say so instead of promising a race it has already lost, and shouldn't let the desk's door-opener scripts lean on "act before August 3."
Fix: Rewrite 09 §3.3/E1 to the post-deadline framing now ("the rules changed August 3 — here's the new triage"), keep the 2027-01 reserve rule as the forward deadline, and scrub "before August 3" urgency from scripts and the E1 sample opener.
M7. The SLA pressure cooker will push NPI into AI tools — the workflow lacks a mandatory redaction gate
What: Intake arrives "by whatever channel the agent already uses" (09 §2.2) — and real agents will text "John and Maria S., 8734 [street], he's 1099 at ~$95k, 612 score…". The response workflow then runs a 25–45-minute analysis block with Claude in the loop (09 §2.3 step 2) under a 4-hour clock. The AI SOP's prohibitions (19 §2.2) are excellent, but the D2 workflow has no explicit step between "raw agent message" and "AI-assisted analysis" — the exact seam where, under time pressure, the paste happens. R16 rates NPI leak likelihood "L"; given this workflow, that's optimistic.
Where: 09 §2.2–2.3; 19 §2.2, §2.10; 24 R16.
Why it matters: One pasted text message = GLBA Safeguards gap, likely fireable per the plan's own words, and it will happen in month 1 on a busy Wednesday, not in a hypothetical.
Fix: Add step 1.5 to the §2.3 workflow: manual normalization onto the intake template (initials/ranges only) before any AI window opens — the normalized form, never the raw message, is the only thing AI ever sees. Put it in the brief SOP checklist and raise R16 likelihood to M.
M8. The DPA lane — the plan's #1 borrower segment — may reduce to "seller credits and gift funds," which is every lender's offer
What: Segment 1 (first-timers stacking DPA) is the plan's largest local opportunity, and every layer of the stack routes around Ready: IHDA requires participation Ready doesn't have (verified); Cook County DPA pairing is unverified; FHLB DPP flows via member banks (Ready isn't one). The honest-education posture is genuinely the right call — but the plan does not model, even roughly, what an honest signpost earns: the Lab's own stop rule (E1 item 11) tolerates zero attributed conversations at day 90, i.e., the plan's stated expected yield from the secondary engine in quarter 1 is ~0 — while file 01 presents the Lab as a co-equal engine built on three timely hooks (one of which expires Aug 3 per M6, and another — the Cook DPA fund — may drain before content ships).
Why it matters: If the manager meeting returns "no Cook-DPA pairing, no DPP channel," product #4 collapses to undifferentiated basics and the first-timer strategy is FHA-plus-honesty against competitors holding $15k of IHDA money. The plan says this would trigger a recut (B5) but doesn't sketch the recut.
Fix: In file 26, add the branch: if the DPA lane fails verification, segment priority reorders to (1) condo/attached-stock triage, (2) tax-arbitrage move-up, (3) first-timers via FHA readiness + honest DPA referral — and the Lab's Asset 1 is repositioned explicitly as a referral-trust play with a stated (small) expected direct yield. Set an honest day-90 content expectation in writing: 0–2 attributed conversations = normal, not failure — and stop selling the Lab as near-term pipeline in file 01.
M9. Personality-fit claims are asserted, not costed — and the week-1 social load lands on the highest-stakes days
What: The calendar's sprint/prep/recovery architecture is genuinely good. But count Version A: 5 outreach sprints + 2 conversation blocks most days + 1 in-person block + Saturday market + open house + monthly 75-minute public workshop ≈ 12–14 hours of direct social exposure weekly, of which weeks 2–4 include ~30 cold dials/week — for a person the brief describes as reserved with no sales history. The plan declares this sustainable (08 scores it 9/10) without any evidence beyond its own design rules, and the burnout guard (23 rule 10) triggers only on ">3 high-social blocks/day," a threshold the schedule is engineered to sit just under.
Why it matters: If the true sustainable dose is lower, the first symptom will be quiet sprint-skipping in week 3 — logged as a volume problem (file 11's "<10 conversations → add one sprint/wk" gate response, which prescribes more of the toxin).
Fix: Make week 2 a deliberate half-dose (waves of 5, not 10) and let week 3 scale only if the week-2 debriefs show energy ≥6/10; change file 11's low-conversation gate response from "add a sprint" to "diagnose energy vs. message first" (the plan already knows how — 23 rule 1 says volume isn't the fix; file 11 contradicts it); log energy-at-close trend as a first-class KPI feeding the Version-B switch.
MINOR FINDINGS
m1. RE/MAX 10 location contradiction. 21 §5 places it in "New Lenox"; 06/15 correctly place it at 15607 S Harlem, Orland Park. Fix file 21.
m2. Hardcoded [DATE_60 = 2026-09-20] (09 §4) violates the plan's own no-hardcoded-dates rule (12 header) — and 2026-09-20 is a Sunday. Make it [WEEK_1_START_DATE]+60 business-day-adjusted.
m3. F-lite start-date contradiction. 09 §4 has 2–3 drills/week running from the start; 11 Phase 2 says "begin Model F-lite prep only" in days 31–60. Pick one (recommend 09's version — drills are cheap) and align.
m4. Unsubscribe-link promise precedes tooling. Script #3 says "the unsubscribe link below works" while [CRM_CHOICE] is unresolved and no list tool exists Day 1 (09 §2.2, 22 §5). For genuinely 1:1 mail, promise behavior ("reply 'stop' and I'll remove you — same day") until a compliant footer/tool exists.
m5. Week-2 calendar has unaccounted gaps (e.g., Wed/Fri 9:00–9:30 missing between anchor hour and sprint) and week-hour totals are ~1h optimistic. Cosmetic, but Mitchell will notice on day 8.
m6. Oak Forest data holes propagate. The unverified Oak Forest ACS row and conflicting attached-share estimates (06 Lens 1/3) leak into E1's audience rationale as load-bearing-looking figures. They're flagged at source; add the flag where reused.
m7. SLA business-hours definitions vary across 09 §2.1 (8:30–5:30) and calendars (7:30 start, inbox by 8:00). Harmless internally; pick one public definition before the SLA is printed.
m8. File 27's "one-page version" says seven asks; C3's fix adds an eighth (runway is Mitchell's own). Renumber when revised.
What genuinely holds up
To be fair to the authors: the verification discipline is the best feature of this plan — the [UNVERIFIED — confirm with manager] regime, the refusal to invent [BPS_COMP], the source log with tier/staleness flags, and the honesty ledger (file 00) are unusually rigorous. The compliance research (TCPA/TSR/CAN-SPAM/815 ILCS 413/1050.940/RESPA §8) is deep, current, and mostly correctly operationalized; the Zillow Flex analysis and its "no" are decisive and well-evidenced; the IHDA-gap honesty (O6 is a model answer); the dual-role year-one rule and decision tree; the AI SOP (toggle kill-list, incident response, LL-2026-04 framing); and the stop-rule instrumentation culture are all real strengths. The failures above are mostly failures of calibration and follow-through (numbers that disagree, deliverables not finished, claims outrunning verification), not of research or intent.
The five most likely ways this plan fails in the first 30 days
Approval starvation (C5). Ready's tiny back office can't produce the disclosure block, product menu, or script approvals inside weeks — the "no calling on unapproved copy" rule (correct) idles the entire ramp calendar, and there is no contingency track or escalation trigger.
The manager meeting returns "unknowns," not answers (C4/C5/M8). No written IL menu, no reviewer for the SLA, no DPA pairing answer — the plan's product spine and its flagship promise are unlaunchable, and no Plan B exists (M1).
The calling channel dies on the DNC/SAN dependency (M4). No Subscription Account Number → under the plan's own rules, cells are untouchable → the sequence degrades to cold email + LinkedIn, and conversation targets become unreachable.
The day-30 gate misfires on miscalibrated targets (C2/M2/M3). A reserved newcomer delivers what the financial model calls base-case (≈10 conversations, 2–4 scenarios) and the 30/60/90 file calls it double-red — triggering demoralization and wrong corrective action (more cold volume) in exactly week 4.
A verbal overreach creates day-one compliance exposure (C4/M5). "Reviewed with my team" without a staffed reviewer, a listing-specific one-pager offer, or live payment math at an open house — each is in the current scripts/calendars, each is the kind of small early slip that a compliance department (or a burned agent) turns into a defining event.
Prioritized fix list for the revision pass
Write the missing deliverables — START_HERE.md, 26_revised_final_plan.md (with a "what changed after red-team" section keyed to these finding numbers), and all three decision_gate/ files. (C1)
Re-base the numbers once, everywhere: relabel file-22 ramps (base→aggressive, conservative→base, new true-conservative 1–3), then recalibrate file-11 gates and file-23 bands to the corrected base. (C2, M2)
Add the runway module: P0-0 runway/benefits questions in file 27, draw-amount + benefits in file 20 Block 4, [RUNWAY_MONTHS] go/no-go rule in R11 and the workbook. (C3)
De-risk the scripts: v0/v1 versions of the review-gate claim; drop Saturday from the published SLA; fix script #7, the forwardable summary, open-house payment math, and O6's hardcoded figures. (C4, M5)
Build the contingency spine: approval-starved holding-pattern track with day-7/14/21 escalations (C5); DNC/SAN fallback channel plan promoted to P0 (M4); employer Plan-B section with walk-away triggers, alternate-sponsor shortlist, and the R20 license-renewal risk (M1).
Fix the acquisition motion for the actual human: sample-brief-first outreach, warm/Crosstown-Hybrid entries ahead of cold corridor slots, week-2 half-dose ramp, and the corrected low-conversation gate response. (M3, M9)
Reframe the condo asset post-deadline and the Lab's honest day-90 yield; sketch the DPA-lane-fails recut. (M6, M8)
Insert the redaction gate (step 1.5) into the Scenario Desk workflow and SOP. (M7)
Sweep the minors (m1–m8) in the same pass — they are an hour of work and several of them (m2, m4) are the kind of detail agents and compliance officers actually notice.