leedab 0.5.3 → 0.6.1

package/context/competitor-landscape.md

@@ -0,0 +1,120 @@
+# Competitor Landscape
+
+Living document. Update monthly or when you spot movement in the market.
+
+## Competitive Zones
+
+LeedAB competes in four overlapping zones. No single competitor covers all four, which is part of the positioning opportunity.
+
+### Zone 1: Cloud Workflow Automation
+
+**Main players:** Zapier, Make (formerly Integromat), n8n, Pipedream, Workato
+
+**What they do:** Connect SaaS tools via APIs. Trigger-action workflows. Marketplace of pre-built integrations.
+
+**How they compare to LeedAB:**
+
+- Strengths: Thousands of integrations already built; huge brand recognition; easy signup; large community.
+- Weaknesses: Require APIs for everything — legacy software without APIs is invisible to them. Data flows through their cloud. Complex workflows need engineering-adjacent skills. They're fundamentally trigger→action tools, not end-to-end operators.
+
+**How LeedAB wins:**
+- Works with legacy ops software (ERPs from the 2000s, regional carrier portals, custom internal tools) that have no APIs
+- Local-first: customer data doesn't flow through a third-party cloud
+- Conversational interface: operators message tasks in plain language instead of building flowcharts
+- Broader scope: a single task can span 5 different apps without requiring an integration for each
+
+**When they win:** Buyer has a well-APIfied SaaS stack, a developer-adjacent workflow owner, and no data sensitivity concerns.
+
+### Zone 2: Enterprise RPA
+
+**Main players:** UiPath, Automation Anywhere, Blue Prism, Microsoft Power Automate Desktop
+
+**What they do:** Desktop automation via recorded or scripted bots. Used in large enterprises for high-volume back-office work.
+
+**How they compare to LeedAB:**
+
+- Strengths: Enterprise sales motion. Compliance certifications. Mature change management tooling. Handle extremely high volume.
+- Weaknesses: Require an engineering team or RPA center of excellence. Implementation projects run months. Bots are brittle — UI changes break them. Priced for enterprise, prohibitive for SMBs.
+
+**How LeedAB wins:**
+- 10-minute setup vs. multi-month deployment projects
+- AI-native: adapts to UI changes instead of breaking
+- SMB pricing
+- Conversational: no bot authoring, no scripts
+- Ops manager can deploy without IT
+
+**When they win:** Enterprise with compliance requirements, high volume, and an existing RPA practice.
+
+### Zone 3: AI Assistants and Copilots
+
+**Main players:** ChatGPT Enterprise, Claude (via Anthropic), Microsoft 365 Copilot, Google Duet
+
+**What they do:** Conversational AI for drafting, summarizing, answering questions. Increasingly adding basic tool use.
+
+**How they compare to LeedAB:**
+
+- Strengths: Massive brand. Network effects (everyone's using them). Fast feature shipping. Huge capability in text/reasoning.
+- Weaknesses: Primarily advisory, not operational — they suggest, the human still has to execute. Data sent to cloud. No persistent workflow capability against the user's specific tools.
+
+**How LeedAB wins:**
+- Takes action end-to-end, not just advice
+- Works with the customer's actual tools (their ERP, their spreadsheets, their portals), not generic document drafting
+- Local-first data handling
+- Persistent memory of the customer's specific ops context
+
+**When they win:** Buyer primarily wants drafting/analysis help, not task execution. Or the buyer is already all-in on Microsoft/Google stack and wants the integrated option.
+
+### Zone 4: Agentic AI Platforms
+
+**Main players:** Lindy, Sierra, Replit agents, Crew AI (developer-focused), Multi-On, browser-automation plays
+
+**What they do:** Build and deploy AI agents for specific tasks, usually via cloud-hosted computer use or API orchestration.
+
+**How they compare to LeedAB:**
+
+- Strengths: Developer mindshare. Fresh VC capital. Polish and iteration speed.
+- Weaknesses: Most are cloud-hosted (misses the local-first value prop). Most are either developer tools or single-purpose agents (customer service, etc.), not general-purpose SMB operators.
+
+**How LeedAB wins:**
+- Local-first positioning is a real differentiator for data-sensitive SMBs
+- Messaging-channel interaction (WhatsApp/Telegram) beats web chat UX for operational teams
+- Horizontal operator use case vs. narrow vertical agents
+
+**When they win:** Buyer is a developer building their own internal agents, or needs a narrow vertical agent (e.g. customer service bot specifically).
+
+### Adjacent: Industry-Specific ERPs and Ops Tools
+
+Examples: industry-specific logistics software (Sennder, Transporeon for freight), HR systems (BambooHR, Rippling), accounting tools (Xero, QuickBooks).
+
+These aren't direct competitors but they ARE where the work currently happens. LeedAB's positioning: we don't replace your ERP, we automate the manual work around it.
+
+## The Headline Pitch vs Each
+
+**vs. Zapier:** "Zapier needs APIs. We don't."
+**vs. UiPath:** "UiPath needs an engineering team. We need 10 minutes."
+**vs. ChatGPT:** "ChatGPT advises. We operate."
+**vs. Lindy/Sierra:** "Our agent runs on your hardware. Yours doesn't."
+
+## What Would Change the Landscape
+
+Watch for these shifts:
+
+- **Zapier ships computer use** — would compress our "no APIs needed" moat. Mitigate by doubling down on local-first and SMB-tailored UX.
+- **Anthropic/OpenAI release polished consumer computer-use products** — compresses the AI capability gap. Our differentiation becomes ops workflow specificity + local-first deployment, not raw AI capability.
+- **A well-funded agentic startup pivots to local-first SMB ops** — most direct threat. Watch Lindy, Sierra, and newer entrants. Our defenses: speed of execution, customer memory/personalization, channel depth, brand.
+- **UiPath launches an SMB offering** — probably too culturally/structurally hard for them, but monitor.
+
+## Battlecard Shortcut (for sales/support use)
+
+When a prospect says "but what about [competitor]?", use this script:
+
+1. **Acknowledge** — "[Competitor] is good at [their actual strength]."
+2. **Shift** — "For operations-heavy SMBs, the challenge is [specific gap]."
+3. **Demonstrate** — "LeedAB does this because [specific mechanism]."
+4. **Proof** — use a specific customer scenario or data point.
+5. **Offer** — low-friction next step (trial, demo, etc.).
+
+---
+
+*Update log:*
+- 2026-04-19: Initial document created

package/context/customer-personas.md

@@ -0,0 +1,91 @@
+# Customer Personas
+
+This file captures who LeedAB serves — specific enough that all agents can tailor output to real buyers.
+
+## Primary ICP: The Overloaded Ops Manager
+
+**Who:** Operations manager or COO at an SMB (10-200 employees), ops-heavy industry.
+
+**Industries:** Logistics, regional distribution, HR/staffing, finance/bookkeeping, professional services (law, consulting), delivery fleets, field services.
+
+**Day in the life:**
+- Starts the day with 40+ emails, half of which are forwarded tasks
+- Bounces between 5-8 apps: email, Excel/Google Sheets, ERP, client portals, accounting software, Slack/WhatsApp
+- Spends 2-4 hours on repetitive data entry, reconciliation, or document generation
+- Fields questions from team members that require pulling data from 2-3 places
+- Stays late to "catch up" on the work that piled up during meetings
+
+**Pains:**
+- Drowning in manual work that clearly should be automated
+- Has tried Zapier / Make / automation tools; hit API walls or complexity walls
+- Can't get engineering resources because the company has 0-2 developers who are already slammed
+- Doesn't trust cloud migration for sensitive data (payroll, contracts, vendor rates)
+- Knows AI could help but doesn't know where to start
+
+**What they want:**
+- Someone (or something) to just do the work
+- Minimal ramp time — they don't have weeks to set up a tool
+- Control over sensitive data
+- Predictable outcomes they can verify
+
+**What they don't want:**
+- Another dashboard to check
+- A new app for the team to learn
+- A month-long implementation project
+- Data in someone else's cloud
+- Hidden costs that scale unexpectedly
+
+**Buying triggers:**
+- New hire didn't work out → "We need to do more with less"
+- Compliance review → "Our data can't be in more clouds"
+- Failed automation project → "We spent $X on Zapier and it didn't work"
+- Scaling breaking → "We can't keep hiring ops people"
+- Mergers/acquisitions → "We have three systems that need to talk"
+
+**How they buy:**
+- Often the champion is not the economic buyer — may need to sell COO or owner
+- Decision timeline: 1-4 weeks from first touch to purchase, often driven by an acute pain event
+- Price sensitivity: high in absolute terms, low relative to the pain if ROI is clear
+- Proof needs: trial is critical; references help; case studies in their industry help a lot
+
+## Secondary ICP: The Owner-Operator
+
+**Who:** Founder/owner of a small business (5-30 employees) who's personally doing ops work.
+
+**Characteristics:**
+- Still doing billing, scheduling, or reconciliation themselves because there's no one else
+- Tech-friendly but not technical (knows how to use software, doesn't write code)
+- Price-sensitive but willing to pay for real time back
+- Wants to grow but is the bottleneck
+
+**Messaging emphasis:** Give yourself your nights and weekends back. Get leverage without hiring.
+
+## Tertiary ICP: The Forward-Leaning CFO/Controller
+
+**Who:** Finance lead at an SMB who's seen the ROI math on ops automation and wants to champion it internally.
+
+**Characteristics:**
+- Understands unit economics and can make the case to ownership
+- Cares deeply about data controls and audit trails
+- Often the one who actually buys (signs the contract) even if the ops manager drives day-to-day use
+
+**Messaging emphasis:** Measurable ops cost reduction. Audit trail. Data stays on-prem. Predictable pricing.
+
+## Anti-Personas (not our customer)
+
+- **Enterprise IT buyer** — wants SOC 2, SSO, SAML, dedicated CSM, procurement cycles. We don't serve them yet.
+- **Individual consumer** — wants a personal assistant. We don't serve them.
+- **Technical founders at tech startups** — will build their own tooling or use direct API access.
+- **Regulated enterprises with complex compliance** — healthcare, financial services at scale. Requires regulatory work we haven't done.
+
+## What Customers Say (Representative Quotes from Site)
+
+- "Payroll used to take two people three days. Now one person sends the files on Telegram and gets the full breakdown back." — HR & Ops, delivery fleet
+- "We get data from three clients in three formats. The agent normalizes everything, flags discrepancies, and gives me one clean sheet." — Ops Coordinator
+- "It runs on a Mac Mini in our office. Vendor rates, payroll, contract details. None of it leaves the building." — COO, regional logistics
+
+Use these quotes verbatim when relevant. Don't fabricate new ones without founder confirmation.
+
+---
+
+*Founders: add real customer names, numbers, and new insights as they emerge. This document gets sharper the more real customer conversations you have.*

package/context/customers/acme-logistics/playbook.md

@@ -0,0 +1,19 @@
+# Acme Logistics — prospect playbook
+
+> Demo prospect. NOT a paying customer. License_key null. Used to validate the multi-tenant flow alongside Tarheel.
+
+## Profile
+
+- **Industry:** logistics (matches Tarheel tier: logistics-pro).
+- **Stage:** prospect — Solutions agent demo target. Mina shares a demo URL during cold conversations.
+- **Hypothetical scale:** 80 deliveries/day, NSW intrastate.
+
+## What this prospect demonstrates
+
+- Per-customer manifest: same tier as Tarheel but smaller agent roster (no people / delivery / onboarding / champion in this demo).
+- Per-prospect playbook: this file gets injected into every agent's system prompt, so a demo conversation feels prospect-specific without the full Tarheel override depth.
+- Conversion path: when this prospect signs, the manifest gets edited to enable the lifecycle agents (delivery / onboarding / champion) and license_key is populated.
+
+## Style + tone
+
+Generic logistics-pro tone — concise, operational. No prospect-specific overrides yet (intentionally).

package/context/customers/bondi-bloom/playbook.md

@@ -0,0 +1,30 @@
+# Bondi Bloom Ceramics — customer playbook
+
+> Demo prospect representing the **maker / e-commerce** SMB segment. Solo-founder + one part-time studio assistant. The customer that's drowning in their own success — Instagram is doing the work but they can't keep up.
+
+## Profile
+
+- **Industry:** retail / handmade ceramics. Sydney-based, primarily online.
+- **Stage:** prospect — Mina would demo this when meeting designer/maker SMBs at markets or design events.
+- **Hypothetical scale:** ~1,200 Instagram followers, 30-50 orders/month, 2 wholesale stockists.
+- **Stack:** Shopify · Instagram · Square (markets) · Xero (manual upload) · ManyChat for IG comments.
+
+## What this prospect demonstrates
+
+- Marketing + brand + content + campaigns all on the rail = a full creative stack for a single founder.
+- Brand override carries the maker voice — not generic e-commerce copy.
+- CX agent handles "where's my order" deflection so Lucy (the founder) can stay at the wheel.
+- No engineering / no devops needed — this manifest shows LeedAB scoping naturally for non-technical SMBs.
+
+## Style + tone
+
+- **Warm, observational.** Brand voice is "studio diary," not "sale ends Friday."
+- **Slow vs fast.** Slow content (origin story, glaze process) builds the audience; fast content (drop alerts, market reminders) converts.
+- **Humble brag.** "Saturday market sold out" reads as gratitude, not flex.
+
+## Operational hooks
+
+- Drop schedule: monthly "kiln drop" Sundays 7pm AEST — primary sales event.
+- Market kit: stock list, float bag, EFTPOS device — Saturday pre-flight checklist lives here.
+- Stockists: never run a site-wide discount that undercuts the wholesale margin.
+- Returns: rarely accept; if a piece arrives broken, replace from the next firing.

package/context/customers/chip-pro/playbook.md

@@ -0,0 +1,107 @@
+# Chip Pro · Customer Playbook
+
+This file gets injected into every agent's system prompt when the chip-pro pack
+is active. It is the customer-specific knowledge that turns generic agents into
+operators who understand this procurement org. Fill in the bracketed fields per
+customer; everything else stays.
+
+---
+
+## Company
+
+- **Customer**: [company name]
+- **Segment**: [fabless designer | IDM | foundry | OSAT | CHIPS Act fab | hyperscaler hardware buyer]
+- **Primary buyer**: [VP/SVP Supply Chain, Strategic Sourcing, Procurement, or Operations]
+- **Reports into**: [COO / BU operations leader]
+- **Pilot scope**: [one workflow first — most often Advanced Packaging Control Tower]
+
+## Foundry / OSAT / Memory / Substrate Suppliers
+
+List the actual counterparties. Replace examples. Names should match how they
+appear in supplier emails and portal exports.
+
+**Foundries**
+- TSMC — [node mix: e.g. N4, N3E, N2]
+- [Samsung / Intel Foundry / GlobalFoundries / UMC / SMIC if applicable]
+
+**OSATs / advanced packaging**
+- [Amkor / ASE / JCET / SPIL]
+- [Package types in scope: CoWoS, InFO, FOPLP, hybrid bonding]
+
+**HBM**
+- [SK hynix / Micron / Samsung]
+- [HBM generation: HBM3e, HBM4]
+
+**Substrates**
+- [Ibiden / Unimicron / Shinko]
+
+## SKU and Program Map
+
+How this customer names things internally. Important because every customer
+spells the same package or wafer family differently.
+
+- SKU naming convention: [example: AI100-N3-CO2 = Accelerator family, N3 wafer, CoWoS Gen 2]
+- Active programs: [program name → BOM root → revenue line]
+- Constrained nodes / packages: [what is currently capacity-locked]
+
+## Commercial Terms
+
+The contract memory the BOM risk and LTB / EOL workflows pull from.
+
+- LTA (long-term agreement) suppliers: [list]
+- NCNR-protected line items: [list]
+- Take-or-pay clauses in effect: [list with thresholds]
+- Capacity reservations / prepayments: [supplier → reserved units, period]
+- Standard payment terms: [Net N from invoice]
+
+## Export Compliance Posture
+
+Inputs the Export Compliance Screener pulls from. Update quarterly or when BIS
+issues a new rule.
+
+- ECCN classifications standardly used: [3A001, 3A090, 5A992, etc.]
+- Foreign Direct Product (FDP) rule exposure: [yes / no per product family]
+- Annex 4 entities to monitor: [if relevant]
+- Restricted destinations (per current rule set): [list]
+- Internal compliance signoff owner: [name + email]
+
+## Internal Tooling and Channels
+
+The agent's "computer use" surface. Update when systems change.
+
+- ERP: [SAP S/4HANA / Oracle / NetSuite] — instance URL, login path
+- Procurement: [SAP Ariba / Coupa / GEP]
+- Planning: [Kinaxis / o9 / SAP IBP]
+- Foundry portals: [list with URLs]
+- OSAT portals: [list with URLs]
+- Mail / chat: [Outlook | Gmail], [Teams | Slack]
+- Audit log destination: [where allocation reports and screening packets get filed]
+
+## Escalation Chains
+
+Who to draft an escalation to for each pain class.
+
+- Allocation gap: [name, role, channel]
+- Compliance hold: [name, role, channel]
+- LTB / EOL: [name, role, channel]
+- Single-source failure: [name, role, channel]
+
+## Standing Rules
+
+Things the agent should always do, or never do. Tighten over time.
+
+- Always: cite source supplier email or portal screenshot when surfacing an allocation change.
+- Always: keep BOMs, supplier emails, and screening packets on the local appliance. Never paste them into the AI provider prompt.
+- Never: send a supplier email without explicit approval.
+- Never: classify ECCN unilaterally. Draft a memo and route to [compliance owner].
+- Audit: every prompt, click, and decision logs to the local audit trail. Inspectable on demand.
+
+## Workflow Cadence
+
+What the agent runs and when.
+
+- Allocation control tower: every weekday morning, 08:00 local. Diff vs prior day.
+- BOM risk monitor: weekly, Monday 09:00 local. Plus event-driven on news / BIS update.
+- Export compliance screener: on-demand per new customer / supplier / PO.
+- LTB / EOL tracker: weekly, plus same-day on any LTB notice received.
+- Supplier follow-up: every 48 hours on open items.

package/context/customers/coastal-built/playbook.md

@@ -0,0 +1,31 @@
+# Coastal Built — customer playbook
+
+> Demo prospect representing the **construction trades** SMB segment. Small residential builder — 1 director (licensed builder), 2 site supervisors, ~7 regular subbies. Pillarix-adjacent; relevant when LeedAB talks to vertical AI players in this space.
+
+## Profile
+
+- **Industry:** construction / residential builds. NSW Central Coast + Northern Beaches.
+- **Stage:** prospect — useful demo for trades SMBs and for partnership conversations with construction-vertical AI players (Pillarix).
+- **Hypothetical scale:** 3-5 active builds, $1.2-2.4M project values, 6-9 month build cycles.
+- **Stack:** Buildxact · Xero · CalTrans for daily site reports · Procore-lite for plan revisions · WhatsApp groups per site.
+
+## What this prospect demonstrates
+
+- Ops + procurement = the load-bearing trades pair. CX matters but is secondary.
+- Compliance covers HBCF (NSW Home Building Compensation Fund), engineer sign-offs, council inspections.
+- The "construction has its own substrate" angle — useful framing for the Pillarix partnership conversation. LeedAB can serve construction without being construction-specific; per-customer manifests handle the vertical depth.
+
+## Style + tone
+
+- **Plain, direct, blue-collar professional.** No corporate fluff.
+- **Time-zoned to AEST/AEDT.** "Tomorrow" means tomorrow on this coast.
+- **Subbie-first.** Communicate via the subbies' preferred channel (WhatsApp group), not email.
+- **Money in writing.** Variations and PC sums get formal change-orders, never just "yeah she'll be right."
+
+## Operational hooks
+
+- Buildxact is single source of truth for job status. Don't act on a verbal that contradicts Buildxact.
+- HBCF certificate must be active before site work.
+- PCD (Practical Completion Date) slips trigger client communication BEFORE they're a slip.
+- Engineer sign-off required for any substitution on engineered specs.
+- Bunnings Trade card limit: $5k per supervisor per week — flag overspend.

package/context/customers/element-pilates/playbook.md

@@ -0,0 +1,33 @@
+# Element Pilates Studios — customer playbook
+
+> Demo prospect representing the **boutique fitness / wellness** SMB segment. 2 studio locations, retention-driven, premium positioning. The customer where every member knows the founder's name.
+
+## Profile
+
+- **Industry:** boutique fitness / wellness — reformer Pilates.
+- **Locations:** Mosman, Manly (Northern Beaches Sydney).
+- **Stage:** prospect — useful demo when talking to studio operators, wellness brands, or community-driven SMBs.
+- **Hypothetical scale:** ~280 active members, 12-class daily schedule across both studios, 8-cap reformer rooms.
+- **Stack:** MindBody · Klaviyo · Squarespace · Square POS · Instagram + Reels.
+
+## What this prospect demonstrates
+
+- CX is the load-bearing agent (member retention is the business model).
+- Marketing + content + campaigns + email round out a community-driven SMB stack.
+- People agent handles instructor scheduling, casual contracts, certifications.
+- Tone override on CX shifts the voice from clinical to "studio family" — illustrates how a single override changes the felt experience of every interaction.
+
+## Style + tone
+
+- **Warm, never transactional.** Members are people, not subscriptions.
+- **Calm.** Pilates voice — measured, embodied, present.
+- **First-name basis.** Always.
+- **Aspiration without pressure.** "Building strength" not "transform your body."
+
+## Operational hooks
+
+- Class capacity: 8 reformers per studio. Waitlist priority is sacred.
+- Monthly community newsletter — first Sunday, ritual cadence.
+- New-member nurture: 30-day sequence with check-in calls at Day 7, Day 21.
+- Casual instructor contracts via Tanda; cert renewals tracked quarterly.
+- Premium positioning: never lead with discount. "First class free" is the only acquisition discount.

package/context/customers/hawkesbury-roofing/playbook.md

@@ -0,0 +1,21 @@
+# Hawkesbury Roofing — prospect playbook
+
+> Demo prospect. Generated by Solutions agent 2026-05-03.
+
+## Brief
+
+/build hawkesbury-roofing, trades — sales pipeline + finance only
+
+## Profile
+
+- **Industry:** trades
+- **Tier:** tier-less
+- **Stage:** prospect — license_key null.
+
+## What this demo shows
+
+6 agents on the rail (coo, concierge, sales, finance, ops, cx). Full Hawkesbury Roofing demo of the LeedAB OS surface.
+
+## Notes
+
+trades — sales pipeline + finance only

package/context/customers/howard-partners/playbook.md

@@ -0,0 +1,33 @@
+# Howard & Partners Accounting — customer playbook
+
+> Demo prospect representing the **professional services** SMB segment. Mid-size accounting firm — 6 partners + 14 staff. Conservative, credentialled, regulated.
+
+## Profile
+
+- **Industry:** professional services — accounting, tax, advisory.
+- **Location:** Newcastle NSW.
+- **Stage:** prospect — Mina would demo this when talking to professional-services SMBs (accountants, lawyers, consultants).
+- **Hypothetical scale:** ~280 active SMB clients, EOFY peak workflow, dual seasonality (June + October).
+- **Stack:** Xero Practice Manager · Microsoft 365 · DocuSign · BGL · Cliniko-equivalent for advisory bookings.
+
+## What this prospect demonstrates
+
+- Compliance + Legal + Finance triad = the load-bearing trio for a regulated SMB.
+- Email override carries the firm-voice rules — no jargon, scope confirmation in writing.
+- People agent handles internal hires + reviews — partnerships have higher people-ops density than equivalent-headcount product cos.
+- Tone shifts conservative across the board — no creative liberty taken on Howard & Partners output without sign-off.
+
+## Style + tone
+
+- **Plain English.** Always. Jargon only when the client uses it first.
+- **Cautious.** Scope confirmations in writing before any substantive advice.
+- **Quietly authoritative.** No exclamation marks. No "we're excited to announce."
+- **Year-round consistency.** EOFY is busy but that doesn't mean voice changes.
+
+## Operational hooks
+
+- TPB registration renewal: tracked in Compliance agent's calendar.
+- PI / Cyber / Mgmt-Liability with Vero — June 30 renewal.
+- AML/CTF reporting cadence for trust-account work.
+- EOFY pre-comms trigger 6 weeks before June 30 + October 31.
+- Never suggest financial products outside accounting authorisation. Refer.

package/context/customers/ledbetter-advisory/playbook.md

@@ -0,0 +1,21 @@
+# Ledbetter Global Advisory — customer playbook
+
+> Provisioned 2026-05-04 via `bin/onboard-customer.sh`. Industry: **professional**. Tier: **null**.
+
+## Profile
+
+- **Industry:** professional
+- **Stage:** prospect (license_key null)
+- **Brain:** `~/leedab-brains/ledbetter-advisory/`
+
+## Style + tone
+
+(replace this stub — what does Ledbetter Global Advisory's voice sound like?)
+
+## Operational hooks
+
+(replace this stub — what's their stack, who's the primary contact, what triggers escalation?)
+
+## Champion
+
+The Champion at Ledbetter Global Advisory hasn't been designated yet. Set this during the kickoff call and update `00_Org/champion-profile.md` in the brain.

package/context/customers/sunset-cafe/playbook.md

@@ -0,0 +1,27 @@
+# Sunset Cafe — prospect playbook
+
+> Demo prospect representing a non-logistics vertical (hospitality). License_key null, tier null — uses manifest agents directly without inheriting from any pack baseline.
+
+## Profile
+
+- **Industry:** Australian hospitality. Independent cafe group, 3 venues (Bondi / Surry Hills / Newtown).
+- **Stage:** prospect — Mina demos this when talking to non-logistics SMBs to show LeedAB isn't industry-locked.
+- **Hypothetical staff:** ~25 baristas + floor + kitchen across 3 sites, 1 ops manager, 2 owners.
+
+## What this prospect demonstrates
+
+- Tier-less manifest: not a logistics-pro install. Manifest carries its own agent roster (coo / concierge / cx / marketing / content / campaigns / people / ops / finance) — distinct from logistics-pro.
+- Industry breadth: shows the Solutions agent flow generalises beyond logistics-pro.
+- Different overrides: cx agent is the load-bearing agent for hospitality (customer concerns, dietary tagging) — gets a tone override that shifts toward "warm, casual" instead of the operational concise tone Tarheel uses.
+
+## Style + tone
+
+- **Warm, casual.** Brand voice is "your friendly neighbourhood cafe" not "logistics ops."
+- **Customer-first.** Frame everything from a guest perspective.
+- **Concrete.** Reference actual menu items / hours / rush times when known.
+
+## Operational hooks
+
+- Brunch rush: 09:00–12:30 weekends, biggest CX volume.
+- Dietary tags: vegan / GF / nut-free are first-class menu metadata.
+- Staff scheduling: roster lives in Tanda; LeedAB doesn't touch that yet.

package/context/customers/tarheel/playbook.md

@@ -0,0 +1,28 @@
+# Tarheel Logistics — customer playbook
+
+> Read this when responding for the Tarheel install. Sets shared context across every agent so cross-agent handoffs don't drop Tarheel-specific facts.
+
+## Profile
+
+- **Industry:** Australian logistics (NSW + interstate freight, ~200 deliveries/day average).
+- **Stage:** Tarheel was LeedAB's first paying customer. Pilot conversion 2026-Q1.
+- **Locations:** primary warehouse Western Sydney; secondary handoff to NZ team via 5am Sydney shift roll.
+- **Stack the customer cares about:** SAP for order management, Cario for TMS. Never login to systems they don't use.
+
+## Style + tone
+
+- **Concise.** No emoji, single-line summaries when possible.
+- **Operational verbs.** "Confirm", "dispatch", "release", "hold" — over softer alternatives.
+- **Time-zoned.** Always quote Sydney time. NZ team gets a parenthetical NZT alongside.
+
+## Hard rules
+
+1. **Cross-customer isolation.** Tarheel data never leaks into another customer's session. Cross-customer references are forbidden.
+2. **Dispatch ID = SAP order ID.** When in doubt, confirm match before acting.
+3. **Coles deductions go through legal.** Never reply informally — it's a formal dispute pattern.
+4. **Customer's primary contact:** Operations Manager. Telegram is preferred 06:00-12:00 Sydney; voice only if urgent.
+5. **Escalation:** anything over $5k impact, anything customer-visible, or anything that touches a contractual SLA — escalate to Mina before action.
+
+## Industry context
+
+Refer to `context/skills/ops-skills.md` and Wiki/customer-use-cases/logistics.md for general logistics patterns (HVNL, Chain of Responsibility, freight classification). Tarheel-specific overrides land here.

package/context/metrics.md

@@ -0,0 +1,74 @@
+# LeedAB Metrics
+
+**Founders: update this file weekly or whenever numbers shift. Agents check this file for current traction data.**
+
+*Last updated: [DATE — fill in]*
+
+## North Star Metrics
+
+| Metric | Current | Last period | Target |
+|--------|---------|-------------|--------|
+| Active paying customers | [TBD] | [TBD] | [TBD] |
+| MRR | [TBD] | [TBD] | [TBD] |
+| Weekly active users | [TBD] | [TBD] | [TBD] |
+| Trial → paid conversion | [TBD] | [TBD] | [TBD] |
+| Net revenue retention | [TBD] | [TBD] | [TBD] |
+
+## Funnel
+
+| Stage | Current period | Conversion |
+|-------|----------------|------------|
+| Website visitors | [TBD] | — |
+| Trial signups | [TBD] | [%] |
+| Activated (completed first task) | [TBD] | [%] |
+| Paying customers | [TBD] | [%] |
+
+## Acquisition Channels
+
+| Channel | Signups | CAC | Notes |
+|---------|---------|-----|-------|
+| Organic / SEO | [TBD] | [TBD] | |
+| LinkedIn | [TBD] | [TBD] | |
+| Twitter/X | [TBD] | [TBD] | |
+| Direct outreach | [TBD] | [TBD] | |
+| Referral | [TBD] | [TBD] | |
+| Paid | [TBD] | [TBD] | |
+
+## Customer Usage
+
+| Metric | Current |
+|--------|---------|
+| Avg tasks per customer per week | [TBD] |
+| Avg active users per customer account | [TBD] |
+| Most common task types | [TBD] |
+| Channels used (WhatsApp / Telegram / Slack / SMS / dashboard %) | [TBD] |
+
+## Retention
+
+| Cohort | M1 | M3 | M6 |
+|--------|----|----|-----|
+| [Month Year] | [%] | [%] | [%] |
+
+## Financial
+
+| Metric | Current |
+|--------|---------|
+| Cash in bank | [TBD] |
+| Monthly burn | [TBD] |
+| Months of runway | [TBD] |
+| Gross margin | [TBD] |
+
+## Qualitative Signal
+
+Recent customer feedback themes:
+- [Theme 1]
+- [Theme 2]
+- [Theme 3]
+
+Recent pain points / churn reasons:
+- [Reason 1]
+- [Reason 2]
+
+## Notes for Agents
+
+When you need a specific number and this file shows [TBD], ask the founders. Don't fabricate metrics for marketing or investor materials — every number that goes out the door needs to be real.