FlatRate Clean Sarasota

Input data (24 months trailing, monthly revenue, seasonal Sarasota market):

Prophet fit with yearly seasonality on, weekly off, 95 percent interval, run against the full 24-month series.

Forecast, next 6 months:

Trend component: underlying growth of roughly $850 per month, driven by Recurring plan signups compounding month over month.

Seasonality component: strong winter peak (snowbird season, November through March) and a summer dip (June through August) tied to fewer Airbnb turnovers and vacationing families pausing Recurring service.

Takeaway for Mike: budget marketing spend AHEAD of the July/August trough (retention offers for Recurring customers) and staff up before October, since the model shows a 24 percent jump from September to November driven by returning snowbirds and holiday move-outs.

Sample pipeline data (last 90 days, 142 leads):

Conversion funnel:

Overall lead-to-recurring-customer rate: 22 percent.

Win rate by lead source:

Bottleneck identified: the Quoted to Booked stage leaks 40 percent of deals, the single biggest drop in the funnel. Sample notes show most losses here cite "comparing quotes" or no response within 48 hours.

Recommendation: the flat-price guarantee is the strongest closer once a lead SEES the number, so the fix is speed, not price. Auto-send the flat quote within 15 minutes of a form fill (currently averaging 3.2 hours) and follow up by text at the 24-hour mark. Realtor referrals and repeat-customer referrals convert 2 to 3x better than paid channels, so reallocate budget: trim Facebook/Nextdoor spend, add a realtor referral incentive ($25 gift card per closed move-out clean).

Source data: q2_2026_metrics.csv (12 rows, weekly bookings/revenue/service-mix by week)

Deck produced (7 slides):

1. Title slide: "FlatRate Clean Sarasota — Q2 2026 Performance Report"
2. Key Metrics Summary table: Total bookings 486, Total revenue $148,700, Average ticket $306, Recurring customers on file 94, Airbnb turnover accounts 37, Realtor accounts 12, Customer satisfaction (guarantee redemption rate) 1.4 percent (very low, meaning the 100 percent guarantee is rarely invoked).
3. Revenue Trend chart (line): weekly revenue climbing from $9,800 (week 1) to $13,900 (week 12), a 42 percent quarter-over-quarter climb tied to Airbnb season ramp.
4. Service Mix chart (bar): Standard Clean 41 percent of bookings, Recurring (weekly/biweekly) 33 percent, Airbnb Turnover 18 percent, Deep Clean 6 percent, Move-Out/Move-In 2 percent.
5. Revenue Mix by Service (pie): Recurring contributes 44 percent of revenue despite being 33 percent of bookings, confirming Recurring as the highest-value segment per booking.
6. New vs Repeat Customer chart (stacked bar): repeat/recurring share of bookings rose from 51 percent in week 1 to 68 percent in week 12.
7. Closing slide: "Next Quarter Focus" bullet: grow Airbnb turnover accounts (highest-margin, lowest-CAC via host referrals) and push more Standard Clean customers into biweekly Recurring plans.

File output: FlatRateClean_Q2_2026_Report.pptx, built with python-pptx, native editable charts and table (not images), ready to open in PowerPoint or Google Slides.

Rough notes in, 11-slide investor deck out:

1. Title: FlatRate Clean Sarasota. Transparent flat-fee cleaning for Sarasota homes and short-term rentals.
2. Problem: Homeowners hate hourly cleaning quotes, they never know the final bill, and cleaners are inconsistent in quality and no-call-no-shows. Airbnb hosts especially need guaranteed same-day turnovers with zero surprises.
3. Solution: One flat price by home size and service type, quoted upfront, no hourly surprises. Vetted, background-checked local cleaners. 100 percent satisfaction guarantee, redo it free or refund it.
4. Product/Service lineup: Standard Clean, Deep Clean, Move-Out/Move-In, Weekly and Biweekly Recurring, Airbnb Turnover (same-day guarantee).
5. Market size: Sarasota-Bradenton MSA has roughly 210,000 owner-occupied households plus 9,400+ active short-term rental listings (AirDNA-class estimate); residential cleaning is a fragmented, low-trust category, ripe for a branded, guarantee-backed operator.
6. Business model: Flat pricing tiers by square footage, recurring subscription revenue as the core engine (highest LTV, lowest CAC over time), 1099 vetted-cleaner network with a managed dispatch layer.
7. Traction: 94 active recurring accounts, 37 Airbnb host accounts, $148,700 quarterly revenue (Q2 2026), 68 percent of bookings now repeat/recurring, 1.4 percent guarantee-redemption rate (strong quality signal).
8. Competition: independent solo cleaners (no guarantee, inconsistent), national franchises like Merry Maids/Molly Maid (hourly pricing, higher cost, less local flexibility), Airbnb-specific apps (no local vetted labor pool). FlatRate wins on price transparency plus local trust.
9. Team: local operator with vetted cleaner network, realtor and property-manager relationships already active in Sarasota.
10. Financial projection: Prophet-modeled forecast shows revenue climbing from $47K/month (summer trough) to $70K/month by December 2026, driven by Recurring compounding growth.
11. The ask: capital or partnership to expand the vetted-cleaner network and replicate the flat-rate, guarantee-backed model in a second Florida Gulf Coast market.

File output: FlatRateClean_Investor_Deck.pptx, professional blue/white color scheme, title + content layouts, ready to present.

Source: Zoom transcript, "FlatRate Clean Weekly Ops Sync," 42 minutes, attendees: Owner, Ops Lead, Scheduling Coordinator.

Meeting Information

• Date: 2026-07-08
• Duration: 42 minutes
• Attendees: Owner, Ops Lead, Scheduling Coordinator
• Type: Weekly Operations Sync

Key Decisions

• Agreed to raise the Deep Clean flat price for homes over 2,500 sq ft by $30, effective next month, to match cleaner time actually spent.
• Decided to make same-day Airbnb turnover confirmation texts mandatory (cleaner must text "done" with a photo of the finished living room) before checkout window closes.
• Agreed to onboard 2 new vetted cleaners this month to cover the growing Recurring waitlist.

Action Items

Discussion Highlights

• Realtor referral channel (Coldwell Banker contact) is outperforming paid ads 2x on win rate; team agreed to formalize a $25 referral thank-you gift.
• Guarantee redemption rate stayed low (1.4 percent) again this month, confirming quality control is working.
• Recurring waitlist now at 14 households, the main reason for hiring 2 more cleaners.

Next Steps: Reconvene next Wednesday to review the new pricing rollout and turnover-photo SOP adoption.

Output file: FlatRateClean_OpsMeeting_2026-07-08.docx, formatted with heading styles and accent tables.

Report generated, key takeaways extracted, saved to vault, and summarized for email delivery.

Top friction points this month

• Quote-to-booking response time averaged 3.2 hours, well above the 15-minute target, causing a 40 percent drop-off at the Quoted stage.
• Two Airbnb turnover jobs ran late because a cleaner's prior job overran, no buffer time was scheduled between back-to-back turnovers.
• Manual price-sheet updates (Deep Clean tier) took two email threads to confirm across scheduling and ops, no single source of truth.

Suggested standing rules going forward

• Auto-send flat-price quotes within 15 minutes of any form fill, no exceptions.
• Build a minimum 45-minute buffer between any two same-day Airbnb turnovers on one cleaner's route.
• Keep ONE master price sheet (single file), all price changes go through it first, no side-channel updates.

Features/practices to try next month

• Text-based "done" photo confirmation from cleaners on every Airbnb turnover, already agreed in ops meeting, formalize it as a required workflow step. Why for this business: hosts are anxious about guest check-in timing, a photo proof text closes that anxiety loop instantly and reduces host-side support calls.
• Realtor referral incentive program ($25 per closed move-out clean). Why for this business: realtor referrals convert at 61 percent versus 28 percent for organic search, this is the cheapest lever to grow revenue.
• Recurring-plan waitlist conversion push: 14 households waiting, hiring 2 cleaners this month should clear most of it and lock in the highest-LTV segment.

Vault note saved: Claude Tips/FlatRate Clean Insights 2026-07-11.md, tagged business-review, flatrate-clean, monthly-insights.

Email summary sent: subject "FlatRate Clean Sarasota, June Insights, response time is costing you bookings," body condenses the three friction points, three rules, and three next-month tries above into a scannable five-bullet email.

Council Plan (per council.py routing)

1. Pricing/positioning question for Airbnb turnover -> google/gemini-3.1-pro-preview [frontend/general, since it's a customer-facing positioning call]
2. Ops feasibility of same-day turnover SLAs -> openai/gpt-5.3-codex [architecture/ops-systems reasoning]

Simulated council responses (format the skill would return and synthesize):

Gemini 3.1 Pro (positioning lens):

"Keep Airbnb turnover as its OWN visible service, not a sub-line of Recurring. Hosts search 'Airbnb cleaning Sarasota' and 'turnover cleaning' as distinct terms from 'house cleaning service.' Burying it under Recurring loses that search intent and the host-specific trust signals (same-day guarantee, linen/restock checklist, damage reporting) that make a host choose FlatRate over a generic maid service. Price it flat per bedroom count, same flat-fee promise as everything else, but market it on its own landing page."

GPT-5.3-Codex (ops lens):

"Turnover cleans have a hard SLA (checkout to next check-in, often under 4 hours) that Standard/Deep Clean bookings do not. That means Airbnb turnover needs its own scheduling lane, not shared crew capacity with Recurring bookings, or you will blow a turnover during a busy weekend. Build the booking flow so turnover jobs get calendar priority and a cleaner buffer window."

Synthesis: Agree with both. Ship Airbnb turnover as its own priced, marketed service line (per Gemini) AND back it with a separate scheduling priority tier (per GPT). Recommendation logged to three-brain-out/log.md: route=council target=airbnb-turnover-positioning outcome=separate-line+priority-sla.

Team config (~/.claude/teams/flatrate-clean-ops/config.json) for a 4-agent team running FlatRate's day-to-day, ready to feed the monitoring dashboard:

{
  "name": "flatrate-clean-ops",
  "members": [
    {"name": "booking-lead", "agentType": "team-lead", "model": "claude-opus-4-8", "isLead": true},
    {"name": "scheduler", "agentType": "general-purpose", "model": "claude-sonnet-4-5"},
    {"name": "cleaner-vetting", "agentType": "general-purpose", "model": "claude-sonnet-4-5"},
    {"name": "review-responder", "agentType": "general-purpose", "model": "claude-haiku-4-5"}
  ]
}

Live task board (kanban) as it would render:

• To Do: "Confirm Saturday Airbnb turnover at 4102 Siesta Dr (checkout 11am, next check-in 3pm)"
• Doing: scheduler -> "Assigning vetted cleaner to Move-Out clean, 3-bed/2-bath, Palmer Ranch"
• Doing: cleaner-vetting -> "Background check pending for new contractor applicant, Sarasota zip 34231"
• Done: review-responder -> "Replied to 5-star Google review, biweekly recurring client, Gulf Gate"

Inbox sample (booking-lead to scheduler):

"Start on task: Saturday has 3 back-to-back Airbnb turnovers within 6 miles of each other. Route the same 2-person crew, confirm buffer time covers drive."

This is the exact single-file Node server + SQLite historical-session pattern from the skill, scoped to FlatRate's real ops load (booking, scheduling, cleaner vetting, review response) instead of a generic demo team.

Session events and routing decisions (as the skill would log them):

1. User: "Build the flat-rate quote calculator for the booking page." -> non-risky path, Claude direct. Built quote-calculator.ts (home size -> flat price lookup: Studio/1BR $99, 2BR $129, 3BR $159, 4BR+ $199 Standard; Deep Clean +$60; Move-Out/In +$80; Airbnb turnover flat $89 regardless of size tier since it's a lighter touch clean).

1. User: "Now check over your work on the quote calculator." -> MUST-FIRE, no self-review. Routed to Codex:

git diff | codex exec --skip-git-repo-check "Review this. Find bugs, risks, missing tests."

Codex findings (simulated per the skill's contract): "Flat price lookup has no floor/ceiling guard, a home size input of 0 or negative silently returns the Studio price instead of rejecting. Airbnb flat fee doesn't scale for a 5-bed vacation rental, same $89 undercuts a Deep Clean-equivalent job. Add a size validator and tier the Airbnb price by bedroom count like the other services." (Routed via three-brain -> Codex review.)

1. User: "The payment integration touches Stripe." -> Risk-path forced review (matches **/Stripe*): announced before running -- "[three-brain] routing to Codex (adversarial-review) -- risk path: Stripe integration" -- ran adversarial review before allowing a 'done' claim on checkout.

1. Output filed to ./three-brain-out/2026-07-11-flatrate-booking/ with codex-review.md and log.md entry: [2026-07-11 09:14] route=risk-review target=stripe-checkout files=3.

Following the Skills Factory template, here is the complete generated skill:

Folder: flatrate-quote-generator/

SKILL.md:

---
name: FlatRate Quote Generator
description: Generates an instant flat-fee cleaning quote for FlatRate Clean Sarasota based on home size, service type, and add-ons. No hourly math, ever.
---

# FlatRate Quote Generator

Produces a transparent, flat-price quote for any FlatRate Clean Sarasota service in seconds, matching the company's no-hourly-billing promise.

## Capabilities
- Calculates flat price by bedroom/bathroom count for Standard, Deep, Move-Out/In, Recurring, and Airbnb Turnover
- Applies recurring discounts (weekly 15% off, biweekly 10% off base Standard rate)
- Flags jobs needing a manual walkthrough (5+ bed, hoarding-level, post-construction)

## How to Use
1. Collect home size (bed/bath) and service type
2. Run `quote.py` with those inputs
3. Present the flat total, no ranges, no 'starting at'

## Input Format
- bedrooms (int), bathrooms (int), service_type (standard|deep|move|recurring_weekly|recurring_biweekly|airbnb)

## Output Format
- A single flat dollar amount plus a one-line breakdown of what's included

## Example Usage
"Quote a 3-bed 2-bath Deep Clean."
"What's the biweekly rate for a 2-bed condo?"

## Scripts
- quote.py: deterministic price-table lookup, no LLM guessing on numbers

## Best Practices
1. Never present a range, FlatRate's differentiator is transparent flat pricing
2. Always state what's included at that tier
3. Route 5+ bedroom or unusual jobs to a human walkthrough, don't guess

## Limitations
- Does not handle custom one-off requests (e.g. exterior window washing) — those need a human quote

quote.py (deterministic, no LLM guessing on the actual dollar figure): a price-table dict keyed by (service_type, bed_count) returning a flat integer, exactly the kind of precision-required calculation the Factory's rules mandate a script for rather than freehand generation.

Round 0 (Fast track selected). Answers walked through per the skill's interview:

1. Who: Solo business owner running a local service business (not a dev), technical level: "comfortable with CLIs but doesn't want to manage servers."
2. Layers selected: Identity, Critical Context, Long-term Knowledge, Promotion. (Skipped Working Memory and Episodic — not needed for a small ops team.)
3. Capture: smart filter (only capture completed bookings, cleaner incidents, and pricing changes, not every chat turn).
4. Format: Markdown + SQLite hybrid (owner wants Obsidian-readable notes but also queryable booking history).
5. Structure: taxonomy (folders: clients/, cleaners/, pricing/, incidents/).
6. Retrieval: hybrid (keyword for cleaner names/addresses, semantic for "what did we do last time this client complained").
7. Injection: project-aware (loads FlatRate context only when working in the FlatRate project folder).
8. Lifecycle: decay + promotion (old booking chatter compresses after 90 days; a cleaner reliability pattern seen 3x, e.g. "always late to Siesta Key jobs," promotes to a permanent flag on that cleaner's profile).

Built artifacts:

• IDENTITY.md: "FlatRate Clean Sarasota. Flat-fee house cleaning, never hourly. Services: Standard, Deep, Move-Out/In, Recurring (weekly/biweekly), Airbnb Turnover. Differentiator: transparent pricing + vetted cleaners + satisfaction guarantee."
• CRITICAL_CONTEXT.md: current focus, open incidents, last pricing change.
• clients/, cleaners/, pricing/, incidents/ taxonomy folders.
• Promotion rule: 3 repeated cleaner complaints on the same address -> auto-flag for reassignment review.

Phase Detection: "Full pipeline for FlatRate Clean Sarasota" -> ALL phases sequentially.

Phase 1 — PROSPECT (dossier): FlatRate Clean Sarasota, local flat-fee residential cleaner. Competitors: Sarasota Maids (hourly, 4.6 stars/210 reviews), The Cleaning Authority Sarasota (flat-rate, 4.4/180), Merry Maids Sarasota (hourly, 4.3/340). Gap found: none of the three lead with an Airbnb-specific turnover offer in their homepage headline, despite Sarasota's dense vacation-rental market (Siesta Key, downtown condos). Search landscape: "house cleaning Sarasota" (~1,300/mo), "Airbnb cleaning Sarasota" (~90/mo, low competition), "move out cleaning Sarasota" (~210/mo). Unclaimed lane: transparent flat pricing marketed specifically to Airbnb hosts.

Phase 2 — PITCH (S.O.U.L.): Strategy: target realtors (move-out referral partnerships) and Airbnb hosts (recurring turnover contracts) as the two highest-LTV segments, not one-off homeowners. Optimization: rank the "Airbnb cleaning Sarasota" term where competitors are weak. Unique Edge: "You'll never get a bill you didn't expect" — flat price shown before booking, always. Loyalty: satisfaction guarantee + same-cleaner-each-time for recurring clients.

Phase 3 — BRAND: Core message: "Know the price before we knock." Visual direction: clean coastal-professional (not sterile-corporate, not beachy-kitsch) — soft blues/whites, real Sarasota home photography, never stock.

Phase 4 — OPERATE: Deploy a 3-agent boardroom: booking/scheduling agent, cleaner-vetting agent, review-response agent (matches the eai-agent-teams config above).

Phase 5 — CAPTURE: This exact Prospect-to-Operate flow, run once for FlatRate, gets spotted as a repeatable pattern for any local flat-fee service business and queued for auto-generation as a reusable skill.

Applying the mandatory editing process to a real FlatRate Clean Sarasota email.

RAW AI DRAFT (rejected):

"In today's fast-paced world, it is important to note that maintaining a clean home can be challenging. We would like to take this opportunity to leverage our team's expertise to provide you with a comprehensive cleaning solution tailored to your unique needs. Please don't hesitate to reach out with any questions."

Failed the read-aloud test: filler opener, buzzword (leverage, comprehensive, tailored), no CTA, no price, doesn't sound like a Sarasota cleaner wrote it.

EDITED FINAL (sent):

Subject: Your flat quote for 123 Siesta Dr, ready to book

Hi Maria,

Here's your quote for a 3BR/2BA Standard Clean at 123 Siesta Dr: $149 flat, every time, no hourly surprises.

Want it recurring? Biweekly drops it to $129 per visit.

All our cleaners are background-checked and local to Sarasota. If anything's off, we send someone back free, no questions, that's our guarantee.

Ready to book? Reply with a day this week or grab a slot here: [link]

Jenna

FlatRate Clean Sarasota

Why this passes: short paragraphs, real numbers, one clear CTA, no corporate filler, sounds like a person named Jenna wrote it, not a bot. Per the skill's mandatory second-pass rule, this got a human accuracy check on the $149/$129 figures against the current price sheet before sending, and a tone check for local voice, not corporate.

Scoring FlatRate Clean Sarasota's candidate workflows on Impact (1-5) and Risk (1-5), applying the 4-question risk check, then reading the verdict matrix.

1. Quote generation from home size/zip

Impact: 4 (directly drives bookings, core differentiator is flat pricing speed). Risk: 2 baseline. Risk check: not binding (still confirmed before booking), not customer-facing in final form, no system dependency, low trust damage if slightly off. No ticks. Risk stays 2.

Verdict: AUTOMATE. Build a rules-based flat-rate calculator (sq ft + room count -> price band), no AI needed at all, this is a spreadsheet problem, not a model problem.

2. Post-clean customer follow-up + review request

Impact: 3 (review volume matters for GMB local ranking). Risk: 3 baseline (customer-facing tick). One tick -> risk floor 3.

Verdict: HYBRID. AI drafts the personalized thank-you + review ask, human (office manager) approves the batch before send, never auto-sent.

3. Airbnb host recurring-schedule confirmation emails

Impact: 2 (routine, low differentiation). Risk: 2, no ticks.

Verdict: MANUAL/light automate. Below the AUTOMATE threshold on impact alone; a simple templated auto-reply covers it, no AI required.

4. Move-out clean scope disputes with realtors (what's included vs damage)

Impact: 4 (protects margin and reputation with realtor referral partners). Risk: 5 (binding: could affect a real estate closing; customer-facing; trust damage severe if wrong). Two ticks -> risk floor 4, and the dispute nature pushes it to 5.

Verdict: MANUAL. This stays fully human. AI is research-assistant only (pulling the original scope-of-work doc for reference), never drafts the resolution.

Loading two constraint files relevant to FlatRate Clean's actual pain points, per the toolkit's "load one or two, matched to the stage" rule.

Constraint 06, Layer Triage, applied: Before building any AI workflow for FlatRate Clean, triage each task: does this need AI, or does a rule/spreadsheet/script solve it better? Flat-rate quoting (home sqft -> price tier) is a lookup table, not an AI task, ship it as a form with conditional logic. Scheduling conflicts (which cleaner, which house, which day) is a constraint-satisfaction problem, a scheduling tool handles it, not a chat model. Where AI genuinely earns its place: turning a raw post-clean photo checklist into a customer-facing "here's what we did" recap message, and drafting Airbnb-host-specific outreach that varies by property type. Reserve the model for judgment and language tasks, not lookups.

Constraint 05, Voice Architecture, applied: Separate FlatRate Clean's tone (direction) from its constraints (testable rules).

Tone: warm, no-nonsense, Florida-casual but competent. Never salesy.

Constraints (testable): no em dashes; always state the flat price in the first two sentences of any quote message; never use the word "affordable" (undersells the flat-rate value prop, use "transparent" or "no surprises" instead); every customer-facing message names the guarantee by name ("100% Satisfaction Guarantee") at least once. These constraints are checkable by a script (grep for banned words, check price appears early) even though the tone isn't. This is the fix for voice drifting generic after 50 AI-drafted messages: rules are enforced mechanically, tone is enforced by spot-check.

Running Phase 1 diagnosis (answers inferred from the business brief) then Phase 2 assembly.

Diagnosis answers: Recurring work = house cleanings booked by size/type (Standard, Deep, Move-Out/In, Recurring, Airbnb turnover). Work arrives via web quote form, phone, and Airbnb-host referral, usually missing exact sqft or access instructions. Between booking and delivery: confirm price tier, assign a vetted cleaner, day-of text confirmation, clean happens, cleaner submits a photo checklist, office reviews, customer gets a completion notice. Scope boundaries: no exterior windows, no carpet shampoo (add-on only), no biohazard, realtors get a written move-out scope so nothing's assumed. Good deliverable: photo-documented, on-time, matches the flat quote exactly, no upsell surprise. Multiple people touch it: intake (office), execution (cleaner), QA (office manager).

Folder structure built:

flatrate-clean-ops/
  CLAUDE.md
  CONTEXT.md
  01_intake/CONTEXT.md, output/
  02_clean/CONTEXT.md, output/ (renamed from generic "process")
  03_review/CONTEXT.md, output/
  04_deliver/CONTEXT.md, output/
  _config/business-rules.md, quality-standards.md, client-context.md
  _templates/quote-template.md, photo-checklist-template.md

business-rules.md carries the real scope lines (no carpet shampoo, no exterior windows, realtor scope-of-work required for move-out). quality-standards.md converts "good deliverable" into checkable items: photo of every room, arrival within 15-min window, price matched quote exactly. Intake stage contract explicitly flags the scope-creep catch: a request for carpet shampoo routes to an add-on quote, not a silent yes.

Phase 1 diagnosis answers (typical for a local service business) and Phase 2 assembly.

Diagnosis: Formats produced at least twice a month: GMB posts, Instagram/Facebook before-after photo posts, a monthly blog post (local SEO), and Airbnb-host email tips. Process: idea -> draft -> photo select -> publish, 4 real steps. Human review point: before publish only, no team big enough for a mid-process gate. Reference material that stays constant: brand voice (Florida-casual, transparent-pricing angle), the 100% satisfaction guarantee wording, service-area list (Sarasota, Siesta Key, Lakewood Ranch, Bradenton edge). "Done" for the most common format (Instagram post): one before/after photo pair, caption under 150 words, price-transparency callout, local hashtag set, ready to schedule.

Structure built:

flatrate-content/
  CLAUDE.md, CONTEXT.md
  01_idea/CONTEXT.md
  02_draft/CONTEXT.md
  03_photo-select/CONTEXT.md
  04_publish/CONTEXT.md
  _config/voice-and-tone.md, format-patterns.md, constraints.md, audience.md

constraints.md starter set: no em dashes, no "synergy/leverage" language, always mention flat pricing or the guarantee once per post, captions under 150 words for Instagram. audience.md splits two reader types: busy Sarasota homeowners (want reliability, price clarity) and Airbnb hosts (want turnover speed and consistency between cleaners). Each draft stage contract asks which audience the piece targets before writing.

FlatRate Clean's closest fit to "multi-stage client delivery" is onboarding a realtor or property-management partner for recurring move-out and turnover referral business, days to weeks, not an hourly job. Running the diagnosis.

Diagnosis: Deliverable: a signed referral agreement plus a working recurring-clean pipeline for that partner's listings/properties. Engagements start with a discovery call (what properties, what typical timeline before showings/check-in, who approves the invoice). Review process: partner sees the scope-of-work doc before the first clean, no drafts otherwise, revisions happen if a property has unusual access or condition. What kills these engagements: unclear scope on move-out cleans (realtor assumes carpet/windows included, they're not) and slow turnaround before a listing photo shoot. After delivery: ongoing recurring relationship, not one-off, so there's a support stage, not a hard handoff. Reusable material: the standard scope-of-work template and the flat-rate tier sheet.

Structure built (adjusted per the skill's guidance for ongoing engagements):

flatrate-partner-eng/
  CLAUDE.md, CONTEXT.md
  01_discovery/CONTEXT.md
  02_scope-agreement/CONTEXT.md, scope-agreement.md
  03_first-clean/CONTEXT.md
  04_review/CONTEXT.md
  05_support/CONTEXT.md  (ongoing recurring relationship, added per partner having no hard end)
  _config/client-brief.md, engagement-terms.md

The scope-agreement stage is called out as the highest-value stage, per the skill: it's the document that prevents the exact failure mode named in diagnosis (realtor assuming carpet/windows are included). Stage 04_review reads only from the approved scope-agreement, never from a verbal understanding.

Applying the raw ICM template's 3-layer routing pattern (map / router / reference) to FlatRate Clean Sarasota's top-level agent workspace, the thing that sits above the ops, content, and client-delivery sub-workspaces already built.

Layer 1, CLAUDE.md (always loaded, the map):

# FlatRate Clean Sarasota — Agent Workspace
Flat-fee house cleaning, Sarasota FL. Standard/Deep/Move-Out-In/Recurring/Airbnb.
Differentiator: transparent flat pricing, vetted local cleaners, 100% guarantee.

## Structure
- flatrate-clean-ops/      -> booking, cleaning, QA, delivery (per-job)
- flatrate-content/        -> marketing content production
- flatrate-partner-eng/    -> realtor/property-manager onboarding (per-partner copy)
- _references/             -> price sheet, service-area list, guarantee wording, brand voice

## Rules
- Flat price always stated in first 2 sentences of customer messages.
- No em dashes. No "affordable", use "transparent"/"no surprises".
- Scope-of-work required before any move-out clean.

Layer 2, CONTEXT.md (the router, 30-50 lines): routes an incoming request to the right sub-workspace by intent, e.g. "new booking" -> ops/01_intake, "write a post" -> content/01_idea, "new realtor partner" -> partner-eng/01_discovery (fresh copy). Names cross-workspace resources every sub-workspace needs: the price-tier sheet and the guarantee wording live in _references/, not duplicated per workspace.

Layer 3, docs/reference: price-tier.md (sqft bands to flat price), service-area.md, guarantee-wording.md, brand-voice.md. Loaded selectively only when a stage's CONTEXT.md names them, never all at once. This is the teaching pattern from the blueprint: an agent dropped into any sub-folder should be able to work without being handed the whole business context, the map plus the local router is enough.

Building the smallest agent-readable structure for FlatRate Clean's owner to run the business knowledge out of, per the kit's four-folder method.

Structure:

flatrate-brain/
  CLAUDE.md
  00_capture/     -> raw customer calls, voice memos, Airbnb host feedback
  10_workflows/   -> frozen recurring processes
  20_reference/   -> condensed decisions and terms
  30_output/      -> finished artifacts (quotes sent, posts published)

Self-audit answers (per template/self-audit.md): Does 00_capture get used weekly? Yes, phone quote requests and Airbnb host texts land there daily, keep it. Does 10_workflows have real frozen processes yet? One so far: "Weekly recurring-clean scheduling" (which days, which cleaner rotation, how to handle a skip request), frozen using the workflow template. 20_reference holds the condensed signal: the flat-price tier decisions (why Deep Clean is priced 40% over Standard, not 25%, decision made after a margin review), the guarantee's exact terms, and past scope disputes resolved (so the same argument doesn't happen twice with a different realtor). 30_output holds sent quotes and published posts, the finished stuff, not drafts.

First frozen workflow (10_workflows/weekly-recurring-scheduling.md): Trigger: Monday morning. Steps: pull the recurring-client list, check for skip/reschedule texts from the weekend, assign cleaners by zip cluster (Siesta Key cleaners stay on Siesta Key routes), send Tuesday's cleaner their day-of list by 6pm Monday. This is the smallest reusable piece worth freezing first, per the kit's "start small" guidance, because it repeats every single week without variation.