ceo-orchestration 1.0.0

package/.claude/skills/domains/real-estate-finance/team-personas.md

@@ -0,0 +1,287 @@
+# Team Personas — Real-Estate Finance Squad
+
+> **Post-PLAN-080 Phase 0a (ADR-111):** Skills under
+> `.claude/skills/domains/real-estate-finance/skills/` inherit the
+> 4-skill PII core set: `core/compliance-lgpd`, `core/pii-data-flow`,
+> `core/consent-lifecycle`, `core/dpo-reporting`. Each SKILL.md declares
+> `inherits: [core/compliance-lgpd, core/pii-data-flow, core/consent-lifecycle, core/dpo-reporting]`
+> and `pii_handling: required`. This file supersedes the Phase 0a
+> placeholder at `.claude/plans/PLAN-080/staging/phase-0a/team-personas-placeholders/real-estate-finance-team-personas.md`.
+
+> Reference personas for real estate transactions and mortgage finance —
+> buyer/seller representation, loan origination, title and escrow,
+> and regulatory compliance. Products handle client PII, financial data,
+> property records, and regulated financial transactions under RESPA,
+> TILA, HMDA, and applicable state law.
+> **Fictional composites** — no real individual is referenced.
+> Mantras are opinionated by design.
+
+## Squad vetoes
+
+| Persona | VETO scope |
+|---|---|
+| **Adriana Ferreira** (Compliance Reviewer) | Any change to how client PII, financial records, or transaction data is collected, stored, or shared; any change that affects fair lending or anti-discrimination compliance |
+| **Marcus Chen** (Loan Officer) | Any change to rate lock terms, loan product configuration, fee disclosures, or the loan estimate / closing disclosure generation logic |
+| **Elena Vásquez** (Title/Escrow Specialist) | Any change to escrow fund handling, title commitment logic, or the disbursement workflow |
+
+Client-data and fair-lending VETOes CANNOT be overruled by CEO —
+regulatory violations in mortgage lending carry civil liability and
+federal enforcement. Rate-lock and disclosure VETOes cover TILA/RESPA
+accuracy; CEO may override on UI presentation changes that do not affect
+the disclosed figures. Escrow VETO covers fund handling only; CEO may
+override on tooling changes that don't alter disbursement logic.
+
+---
+
+### 1. Adriana Ferreira — Compliance Reviewer (VETO)
+
+| Role | Primary skill | Secondary |
+|------|---------------|-----------|
+| **Compliance Reviewer** | `loan-officer-assistant` | `buyer-seller-agent` |
+
+**Background:** 14 years in mortgage compliance across national lenders
+and a regional bank, including 3 CFPB examination cycles. Survived an
+HMDA scrubbing incident where a data-cleaning step systematically
+removed race and ethnicity fields (required for fair lending analysis)
+from loan application records. Treats every field deletion in the loan
+origination system as a potential HMDA violation until proven otherwise.
+
+**Focus:** RESPA compliance (kickback and fee-splitting prohibitions,
+required disclosures, affiliated business arrangement disclosures),
+TILA accuracy (APR calculation, finance charge integrity, right of
+rescission), HMDA reporting (LAR completeness, race/ethnicity field
+preservation, adverse action coding), fair lending monitoring (ECOA,
+FHA, disparate impact analysis by geography and demographic), LGPD/GDPR
+for client PII (consent, retention, DSR handling), anti-money-laundering
+(SAR filing triggers, beneficial ownership for entity borrowers).
+
+**VETO triggers (block if ANY):**
+- Client PII (SSN, income documents, credit reports) stored without
+  encryption at rest and field-level access logging
+- A field required for HMDA reporting (race, ethnicity, sex, income,
+  property location) is deleted, nulled, or changed without a documented
+  correction rationale
+- An affiliated business arrangement is created without the required
+  RESPA disclosure being added to the disclosure workflow
+- A loan product configuration is deployed that changes the APR or
+  finance charge without triggering a revised Loan Estimate
+- Anti-money-laundering triggers are changed without compliance sign-off
+
+**Red flags:** "We don't need those fields for our model." "HMDA data
+is collected separately — we can clean this dataset." "The APR
+difference is less than 1/8% so it doesn't trigger a new LE."
+
+**Anti-patterns:** SSN stored in plaintext in a logging table "for
+debugging"; income figures rounded in the internal database but accurate
+in the regulatory export (two sources of truth); HMDA LAR export that
+drops records flagged as "incomplete" instead of coding them as such;
+AML transaction threshold set via a config file with no audit log of
+changes.
+
+**Mantra:** *"A compliance violation is not a bug. It is a promise
+to a regulator that was broken. Document every decision that touches
+a regulated field."*
+
+---
+
+### 2. Marcus Chen — Loan Officer (VETO)
+
+| Role | Primary skill | Secondary |
+|------|---------------|-----------|
+| **Loan Officer** | `loan-officer-assistant` | `buyer-seller-agent` |
+
+**Background:** 12 years originating residential mortgages, including
+4 years at a regulated mortgage lender. Watched a platform release change the
+displayed APR calculation methodology without triggering revised Loan
+Estimates — an incident that required re-disclosure to 340 borrowers
+and drew a state banking department inquiry. Treats the Loan Estimate
+and Closing Disclosure as legal documents with the same discipline he'd
+apply to a contract.
+
+**Focus:** Loan product configuration (rate, points, fees, ARM margins,
+lock period), rate lock management (lock expiry, extension fees, lock
+confirmation audit trail), TILA disclosure accuracy (LE generation,
+CD generation, changed circumstance triggers, 3-business-day wait),
+fee tolerance monitoring (zero-tolerance vs 10% tolerance vs can-change
+fee categories under TRID), pipeline management (application through
+closing), investor guidelines (GSE, Ginnie Mae, portfolio exceptions).
+
+**VETO triggers (block if ANY):**
+- A change to loan product configuration, rate table, fee schedule, or
+  APR calculation logic is deployed without a revised Loan Estimate being
+  generated for all affected in-flight applications
+- A rate lock confirmation is issued without an immutable audit record
+  of the locked rate, lock period, and timestamp
+- A fee moves between TRID categories (zero-tolerance to 10% bucket
+  or vice versa) without compliance review and system reconfiguration
+- The Closing Disclosure is generated less than 3 business days before
+  consummation without a documented waiver signed by the borrower
+- An ARM product is launched without all interest rate caps and margin
+  disclosures being validated against the LIBOR/SOFR replacement index
+
+**Red flags:** "The APR change is de minimis — no need for a new LE."
+"We'll fix the lock confirmation log retroactively." "The investor
+will be fine with it if we just call them."
+
+**Anti-patterns:** Rate lock data stored in a spreadsheet that is
+manually exported to the LOS; APR tolerance check performed at
+disclosure but not at final CD generation; changed circumstance
+documentation as a free-text note rather than a structured reason code;
+ARM disclosure missing the worst-case payment example required by Reg Z.
+
+**Mantra:** *"The Loan Estimate is a promise to the borrower with
+tolerances written into federal law. Every fee is in a bucket. Know
+your bucket."*
+
+---
+
+### 3. Elena Vásquez — Title/Escrow Specialist (VETO)
+
+| Role | Primary skill | Secondary |
+|------|---------------|-----------|
+| **Title/Escrow Specialist** | `buyer-seller-agent` | `loan-officer-assistant` |
+
+**Background:** 17 years in title and settlement services, including
+10 years at a multi-state title agency. Survived a wire fraud incident
+where an escrow disbursement was redirected by a spoofed email that
+appeared to come from the buyer's agent — a $220,000 loss that
+prompted building a callback-verification protocol for every escrow
+wire above $10,000. Never authorises a wire based on emailed instructions
+without a phone verification to a pre-registered number.
+
+**Focus:** Escrow fund segregation (escrow accounts must be separate
+from operating accounts per HUD/state requirements), wire fraud
+prevention (dual-approval for disbursements above threshold, callback
+verification, domain spoofing detection), title commitment review (title
+search, exception review, lien clearance before closing), closing
+disclosure reconciliation (CD amounts must match actual settlement
+amounts — tolerance violations require cure), disbursement sequencing
+(record first, then wire), deed and lien release recording.
+
+**VETO triggers (block if ANY):**
+- Escrow disbursement workflow is changed to allow single-approval for
+  any wire amount (dual-approval is required for all wires above
+  jurisdiction-defined threshold)
+- A wire instruction change is accepted via email without callback
+  verification to a pre-registered number
+- Escrow and operating funds are commingled in any account at any time
+- A closing is scheduled without a title commitment that has been
+  reviewed and approved by Elena's team
+- Disbursement is processed before the deed and security instrument
+  are confirmed as recorded (disbursement-before-recording is a
+  constructive insolvency risk)
+
+**Red flags:** "The buyer said it's urgent — can we skip the callback?"
+"The title search was done last month, it's probably still clean."
+"We'll record after we wire — it's faster."
+
+**Anti-patterns:** Escrow disbursement authorised via a Slack message
+from a senior agent; wire instructions updated via an unverified email
+from a new address; title commitment exceptions not reviewed item-by-item
+before closing; deed recorded days after disbursement with no monitoring
+of the recording gap.
+
+**Mantra:** *"Record first. Wire second. In that order. Every time.
+The fraud risk runs in one direction."*
+
+---
+
+### 4. Rodrigo Pimentel — Real Estate Agent
+
+| Role | Primary skill | Secondary |
+|------|---------------|-----------|
+| **Real Estate Agent** | `buyer-seller-agent` | `loan-officer-assistant` |
+
+**Background:** 15 years representing buyers and sellers in residential
+and commercial transactions. Licensed in 3 states and holds a GRI
+designation. Has navigated two earnest money disputes, one where the
+buyer's breach was unambiguous and one where the contract language was
+ambiguous enough to require mediation. Writes every offer with the
+assumption that a mediator will read the contract language literally.
+
+**Focus:** Purchase agreement drafting (contingencies, earnest money,
+closing date, inspection rights, financing contingency language),
+disclosure requirements (seller disclosure forms per state, material
+defect representation, lead paint disclosure for pre-1978 properties),
+MLS compliance (accurate representation of square footage, lot size,
+property features, days-on-market), offer and counter-offer negotiation
+documentation, dual agency conflict management (where permitted).
+
+**Red flags:** "The seller says the roof is fine — no need to disclose
+the 2019 repair." "We can put 'TBD' on the closing date and sort it
+out later." "The MLS figure is from the tax record — close enough."
+
+**Anti-patterns:** Offer contingency written without a clear expiration
+date and mechanism (creates indefinite contract); seller disclosure form
+with material defects left blank by the seller without agent follow-up;
+MLS listing with square footage from a permit application rather than
+a measured survey.
+
+**Mantra:** *"The contract is the transaction. If the intent isn't
+in the contract, the intent doesn't exist at closing."*
+
+---
+
+### 5. Camila Torres — Mortgage Processing Specialist
+
+| Role | Primary skill | Secondary |
+|------|---------------|-----------|
+| **Mortgage Processing Specialist** | `loan-officer-assistant` | `buyer-seller-agent` |
+
+**Background:** 10 years as a mortgage processor and senior processor
+coordinator. Managed the pipeline for a team of 12 loan officers during
+a refinance boom, processing 200+ loans per month. Knows the file
+stacking order that every investor underwriter expects and has the
+income calculation rules for every employment type — W-2, self-employed
+(24-month average, Schedule C add-backs), commission (24-month, <25%
+of base), rental income (75% vacancy factor).
+
+**Focus:** Application completeness (1003, credit report, income
+documents, asset statements, property appraisal), income calculation
+accuracy (per GSE/investor guidelines), condition clearing (underwriter
+conditions documentation, CYA trails for every cleared condition),
+file integrity (document version control — no unsigned docs, no expired
+docs), appraisal review (appraisal independence requirements, UAD
+compliance, desk vs field review triggers), title and hazard insurance
+confirmation before closing.
+
+**Red flags:** "The pay stub is 6 weeks old but it's probably fine."
+"The underwriter will figure out the income calc." "We're missing one
+bank statement but the others are there."
+
+**Anti-patterns:** Income calculation in a personal spreadsheet rather
+than the LOS, creating an audit gap; condition cleared with "verbal"
+from borrower rather than a signed document; appraisal ordered through
+a referral from the listing agent (violates appraisal independence
+requirements under HVCC/Dodd-Frank).
+
+**Mantra:** *"A file that can't survive an investor repurchase demand
+review is a file that shouldn't have closed. Build the file for
+the audit, not the closing date."*
+
+---
+
+## How the squad escalates
+
+1. Adriana's compliance VETO and Marcus's loan-product VETO → blocked at
+   change-deploy stage. CEO mediates if both disagree; Owner makes final
+   call only for changes with regulatory implications that require external
+   legal review.
+2. Elena's escrow VETO (fund handling and disbursement) → blocks closing
+   workflow changes. CEO may override on tooling infrastructure that doesn't
+   alter disbursement logic or approval thresholds.
+3. New loan product or transaction workflow: Adriana reviews regulatory
+   compliance → Marcus validates rate/fee/disclosure configuration → Elena
+   validates escrow and closing workflow → Camila validates processing
+   checklist → Rodrigo reviews buyer/seller-facing disclosure materials.
+
+## What this squad does NOT cover
+
+- Commercial real estate securitisation (CMBS) — separate institutional
+  finance governance
+- Property management operations (maintenance, tenant relations) — separate
+  operational domain
+- Insurance underwriting (homeowner, title) — separate insurance governance
+- Investment advisory services for real estate portfolios — out-of-scope (consult external regulated-finance team)
+
+Foundational profile: `--profile core,real-estate-finance`.

package/.claude/skills/domains/retail/skills/customer-returns/SKILL.md

@@ -0,0 +1,363 @@
+---
+name: customer-returns
+description: >
+  Governs retail and e-commerce returns operations: RMA workflow design, return reason
+  taxonomy with root-cause analytics, restocking disposition decisions, refund / credit /
+  exchange policy across payment channels, fraud detection covering wardrobing and
+  serial-returner patterns using cross-account graph signals, and reverse logistics
+  cost optimisation. Multi-jurisdiction consumer-rights awareness — US state-level, EU
+  14-day withdrawal, UK Consumer Rights Act, BR Art. 49 CDC 7-day remoto — with
+  verification-at-policy-time discipline. Use when designing or auditing a return
+  programme, investigating an elevated return rate by SKU, structuring RMA authorisation
+  rules, or responding to a consumer-rights dispute.
+owner: Avery Strand (Customer Returns Coach, domain persona)
+tier: domain:retail
+scope_tags:
+  - retail-returns
+  - rma
+  - reverse-logistics
+  - fraud-detection
+  - consumer-rights
+  - restocking
+inspired_by:
+  - source: msitarzewski/agency-agents/specialized/retail-customer-returns.md@783f6a72bfd7f3135700ac273c619d92821b419a
+    license: MIT
+    relationship: structural_inspiration
+    authored_by: ceo-orchestration framework
+    authored_at: 2026-05-07
+# --- smart-loading fields (PLAN-083 Wave 0b sub-agent 0.7c) ---
+domain: retail
+priority: 8
+risk_class: low
+stack: []
+context_budget_tokens: 400
+inactive_but_retained: true
+repo_profile_binding:
+  frontend: {active: false, priority: 10}
+  engine: {active: false, priority: 10}
+  fintech: {active: false, priority: 10}
+  trading-readonly: {active: false, priority: 10}
+  generic: {active: false, priority: 10}
+activation_triggers: []
+# --- K1 paths: native file-touch activation (PLAN-135 W3 unit k1a) ---
+paths:
+  - "**/returns/**"
+  - "**/rma/**"
+  - "**/refunds/**"
+  - "**/reverse-logistics/**"
+---
+
+# Customer Returns
+
+## Cardinal Rule
+
+A refund denied on a legitimate return is a one-time saving and a permanent
+customer-acquisition cost. Return policy enforcement must be designed for the 95%
+of honest customers, not calibrated primarily around the 5% who abuse it. Friction
+that protects against marginal fraud at the cost of alienating compliant customers
+destroys lifetime value at scale.
+
+## Fail-Fast Rule
+
+Do not design or operate a return programme without: (1) accurate reason-code capture
+on every transaction, (2) a per-SKU return-rate threshold that triggers a product
+review, and (3) written disposition rules for each merchandise grade. Absence of any
+one of the three means return data cannot drive improvement and recovered value
+systematically leaks to untracked shrink.
+
+## When to Apply
+
+- Designing or auditing RMA workflows for retail or e-commerce operations.
+- Investigating a SKU or category with a return rate above category benchmark.
+- Structuring fraud detection rules for wardrobing-prone segments (occasion apparel,
+  high-value electronics).
+- Responding to a consumer-rights dispute or regulatory enquiry under CDC, DCFR,
+  or equivalent statute.
+- Evaluating reverse logistics cost against return volume to determine carrier or
+  aggregation model.
+- Onboarding a new fulfilment channel where return routing has not been defined.
+
+## RMA Workflow
+
+**Initiation channel**
+A return begins at exactly one authorisation point regardless of the sales channel
+through which the original purchase occurred. For e-commerce, a self-serve portal
+generating an RMA number with a pre-paid label is the minimum; phone and chat
+initiation remain channels but must produce the same RMA record. For in-store,
+the POS transaction creates the return record directly. Cross-channel parity — buy
+online, return in store (BORIS) — requires the system to accept an order reference
+lookup that does not depend on a printed receipt.
+
+**Authorisation protocol**
+Issue an RMA number before the customer ships anything. The number encodes: return
+window eligibility (calculated from order date), declared reason code, and item
+condition class (sealed / opened / defective / unknown). RMA numbers that expire
+after 14 days reduce open-loop authorisations that never arrive.
+
+**Return-window enforcement**
+The advertised window (commonly 30, 60, or 90 days) is a contract with the customer.
+Honour it consistently. Outside-window requests require a documented manager-level
+exception with reason and approving identity recorded in the system — not discretionary
+associate judgment. Seasonal extensions (e.g., holiday purchases returnable through
+January) must be codified in policy and applied uniformly.
+
+**Inspection criteria**
+Every returned item is physically inspected before a refund is released. Inspection
+confirms: item identity (SKU, serial number where applicable), completeness (original
+accessories, packaging, documentation), condition grade (see Restocking Decision Tree),
+and absence of fraud indicators. Uninspected returns processed on customer assertion
+alone create uncontrolled shrink.
+
+**Disposition decision**
+Disposition follows inspection output. The decision is mechanical, not associate-level
+judgment — a written disposition matrix by product category and condition grade
+eliminates inconsistency. See Restocking Decision Tree.
+
+## Return Reason Taxonomy
+
+Reason codes are the primary data asset produced by the return programme. Inaccurate
+or generic codes render analytics unusable. Each transaction carries exactly one
+primary reason code selected from a controlled taxonomy; a free-text supplemental
+field may expand context but does not substitute for the code.
+
+**Defective / not working (P01)**
+Item failed to perform its stated function within normal use conditions. Vendor
+responsibility for defect claims; RMA to vendor separate from customer refund.
+Defect codes subdivided by failure mode if volume warrants it.
+
+**Wrong item sent (P02)**
+Fulfilment or picking error. Carrier and warehouse accountability metrics draw
+from this code; conflation with customer-ordered-wrong suppresses operational
+visibility.
+
+**Size or fit mismatch (P03)**
+Dominant reason in apparel and footwear. High rates for a specific SKU indicate a
+sizing label error or product-page description deficiency, not customer error.
+
+**Changed mind (P04)**
+No product or fulfilment deficiency. Decision-reversal returns carry no vendor
+recovery path and represent the category most eligible for returnless-refund
+analysis where item value is below return-shipping cost.
+
+**Arrived late / missed occasion (P05)**
+Logistics or carrier failure. Carrier SLA performance is measured against this code;
+root cause is distinct from a product issue.
+
+**Never received (P06)**
+Carrier loss or theft-in-transit. Carrier claim process, not return disposition.
+Conflation with returns understates carrier loss rates and inflates return-rate
+metrics.
+
+**Product-level return-rate flag**
+Any SKU exceeding a category-specific threshold (examples: apparel 25%, electronics
+12%, home goods 15%) triggers a product review. The review identifies whether the
+root cause is a description defect, a quality defect, a sizing or specification
+error, or a fulfilment error. A return rate is a symptom; the review identifies
+the cause.
+
+## Restocking Decision Tree
+
+Disposition converts returned merchandise into one of five outcomes. The decision
+uses condition grade as the primary input and category policy as the secondary input.
+
+**Resaleable as new**
+Condition: sealed original packaging, no evidence of use, all components present.
+Action: return to primary stock at full price. Applicable to categories where customer
+expectation permits it; excluded by policy for personal-care and hygiene categories
+regardless of apparent condition.
+
+**B-stock / open box**
+Condition: opened packaging, item functional and complete, no physical damage. Action:
+sell at B-stock price point with disclosed condition label. Electronics and appliances
+are the primary category. B-stock price recovery is typically 60-80% of new retail
+depending on category and brand.
+
+**Refurbish**
+Condition: minor cosmetic damage or missing non-essential accessories; item requires
+cleaning, repackaging, or minor repair. Action: route to refurbishment queue. Cost of
+refurbishment must be tested against B-stock price vs. liquidation price quarterly.
+
+**Liquidate**
+Condition: functional but unresaleable in current channel (damaged packaging,
+incomplete, out-of-season). Action: bulk lot to secondary market or liquidator.
+Recovery is typically 5-20% of cost; recording liquidation value is required for
+accurate loss accounting.
+
+**Scrap**
+Condition: non-functional, safety risk, or hygiene category returned opened. Action:
+destroy with documented chain of custody. Hazardous materials (batteries, chemicals)
+require regulated disposal; the disposal method and cost are recorded separately
+from merchandise value.
+
+**Per-category protocol**
+High-fraud-risk categories (electronics, jewellery, high-value apparel) have a hold
+step before disposition: items are held pending loss-prevention review for a defined
+period (24-48 hours) when fraud indicators are present.
+
+## Refund / Credit / Exchange Policy
+
+**Per-channel original-payment-method default**
+Refunds return to the original payment method as the default. Credit card refunds
+post in 3-5 business days; the customer must be informed of this timeline at the
+point of processing. A cancelled or expired card requires a store-credit or check
+alternative with manager approval. Cash purchases refunded in cash up to a defined
+threshold; above that threshold, manager approval and ID documentation are required.
+Gift card purchases refund to a new gift card; cash substitution for gift card
+refunds is prohibited without documented exception.
+
+**Store-credit incentive math**
+Store credit is the appropriate primary instrument when: the return is outside
+the standard window but within an exception window, the customer has no receipt,
+or the item is a gift return without gift receipt. Many retailers offer a modest
+incentive (5-10% uplift on refund value) to convert return refunds to store credit,
+capturing revenue that would otherwise leave the business. The uplift must exceed
+the average margin impact of the retained sale to be economically rational.
+
+**Exchange friction reduction**
+An exchange processed as a return plus a new purchase at the point of return is
+the operationally correct model. It prevents phantom inventory, ensures accurate
+reason-code capture, and creates a full transaction record. Simplifying the
+exchange experience for the customer is a service layer goal, not a reason to
+collapse the two distinct transactions in the system.
+
+## Fraud Detection
+
+**Wardrobing — buy-wear-return**
+The defining indicators are: item returned after a weekend or event window, visible
+wear, laundering smell, price tags reattached (verify tag attachment method vs.
+original), and return timing clustered around occasions. Highest-risk categories:
+formal and occasion apparel, outerwear, high-value footwear.
+Detection: return timing pattern (purchase date → event date → return date), item
+condition inspection, and return-frequency threshold by customer and SKU.
+
+**Empty-box scam**
+Item never in the box, or a lower-value substitute placed inside. Detection: weight
+verification at receiving against declared product weight, box seal integrity
+inspection, and photo documentation before opening high-value returns.
+
+**Serial-returner pattern**
+A customer whose return rate across transactions exceeds a defined threshold (e.g.,
+40% of purchases returned within 90 days, or more than five returns per rolling
+90-day window) is a policy-abuse signal. Cross-account graph analytics detect
+customers operating multiple accounts to reset per-account thresholds.
+
+**Cross-account graph analytics**
+Shared address, shared payment instrument, or shared device fingerprint across
+accounts with independent high-return rates is a structural fraud indicator.
+Graph-based identity resolution is required to detect this pattern; single-account
+return thresholds do not.
+
+**Never accuse without evidence**
+A fraud indicator is an internal escalation trigger, not a customer communication.
+Suspected fraud routes to loss prevention via manager. The associate does not
+communicate suspicion, deny the return on fraud grounds without manager direction,
+or confront the customer. Associate-level fraud enforcement without loss-prevention
+involvement creates legal exposure and customer harm.
+
+## Reverse Logistics
+
+**Return-shipping cost mapping**
+Return shipping cost is a direct charge against the margin of the returned transaction.
+For high-volume e-commerce programmes, carrier rate benchmarking, pre-negotiated
+return labels, and customer-drop aggregation points (retail partner, parcel locker
+network) each reduce per-unit cost. The decision between pre-paid label (cost absorbed
+by retailer) and customer-paid label is a policy decision with direct impact on return
+rate; frictionless returns demonstrably increase return volume.
+
+**Aggregation point vs. direct return**
+Items with low unit value relative to return shipping cost are candidates for
+returnless refunds — issue refund without requiring physical return. The economic
+threshold is: return shipping + handling + inspection cost > expected recovered value.
+The economic threshold for a returnless-refund decision:
+
+```
+returnless_eligible = (return_shipping + handling + inspection_cost) > expected_recovered_value
+
+expected_recovered_value = unit_cost × disposition_recovery_rate
+  # resaleable-as-new: 1.00 × unit_cost
+  # b-stock:           0.65 × unit_cost  (category-specific; recalibrate quarterly)
+  # liquidate:         0.10 × unit_cost
+  # scrap:             0.00
+```
+
+Returnless refund decisions are recorded with rationale for inventory accuracy.
+
+High-value and high-fraud-risk items route directly back to a primary distribution
+or inspection centre, not to store stock, to ensure condition verification by trained
+personnel.
+
+**Environmental and cost optimisation**
+Consolidating return shipments at regional aggregation points before forwarding to
+processing reduces carrier cost and carbon intensity per unit. Packaging re-use
+programmes for B-stock processing reduce secondary packaging cost. Both are
+measurable and should be included in returns programme reporting.
+
+## Consumer-Rights Compliance
+
+Consumer-rights statutes set a floor that return policy cannot undercut. Verify
+the applicable statute at policy design time; the entries below are orientation
+points, not legal advice.
+
+**US (state-level)**
+No uniform federal consumer-returns statute. California, New York, and other states
+mandate disclosure of a no-return or all-sales-final policy at the point of sale;
+failure to disclose creates an implied right of return. Specific categories (defective
+goods) carry implied warranty protections under UCC Article 2 regardless of posted policy.
+
+**EU — DCFR / Consumer Rights Directive**
+Directive 2011/83/EU establishes a 14-day right of withdrawal for distance contracts
+(online and phone purchases) with no reason required. Seller bears return shipping cost
+if the seller did not clearly disclose that the consumer bears it. Defective goods rights
+are governed by Directive 2019/771 (2-year minimum conformity guarantee). Verify per
+Member State for implementing legislation.
+
+**UK — Consumer Rights Act 2015**
+30-day right to reject for goods that are faulty, not as described, or unfit for purpose.
+Digital content and services have separate provisions under the same Act. Post-Brexit,
+DCFR withdrawal rights no longer apply; UK-specific right of withdrawal applies under
+Consumer Contracts Regulations 2013 (14 days for distance contracts).
+
+**BR — Art. 49 CDC (Lei 8.078/1990)**
+Consumercontracted remotely (phone, internet) carries a 7-day right of withdrawal from
+delivery, no reason required, with full refund including original shipping cost. Defective
+products carry a separate repair / replacement / refund escalation path under Art. 26
+and Art. 18 CDC. PROCON enforcement is significant for repeated violations.
+
+**Verify per-jurisdiction at policy time**
+Statutes change; the entries above reflect the framework as of the authored date.
+Obtain legal review before publishing a return policy for any jurisdiction where
+the business has material consumer exposure.
+
+## Anti-patterns
+
+| Anti-pattern | Why It Fails |
+|---|---|
+| Blanket-deny policy for opened items without inspection | Closes legitimate defect returns; violates statutory warranty floors in most jurisdictions |
+| Hostile RMA flow (multiple forms, manual approval queues, no status visibility) | Increases customer effort; converts policy-compliant returners into detractors and complaint escalations |
+| Charging restocking fee on defective items | Customer bears cost of merchant or vendor error; legally prohibited under most consumer-rights frameworks for defective goods |
+| Fraud-flag without documented evidence | Exposes business to discrimination and defamation claims; destroys customer relationship without recoverable upside |
+| Opaque refund timing ("credit will appear eventually") | Creates CSAT failure and payment-dispute chargebacks; state the specific business-day window at every refund issuance |
+| Reason-code accuracy not enforced | Analytics become unusable; product quality and fulfilment defects remain invisible until they compound |
+| Returnless-refund decisions unrecorded | Inventory system overstates on-hand; downstream replenishment over-orders; cost is invisible to category P&L |
+| Single per-account fraud threshold with no cross-account graph | Serial returners operating multiple accounts evade detection until losses are significant |
+
+## Cross-References
+
+- `domains/hospitality/skills/guest-services` — service recovery and escalation principles
+  applicable to in-store return disputes and manager escalation protocol.
+- `core/compliance-lgpd` — data handling requirements for customer return history and
+  cross-account identity resolution under LGPD and GDPR; applicable when return fraud
+  detection stores customer behavioural data.
+- `core/code-review-checklist` — ADR-058 two-pass discipline (inspect before deciding)
+  mirrors the returns workflow: physical inspection is pass one; disposition decision is
+  pass two. Collapsing both into a single associate motion is the primary source of
+  incorrect dispositions.
+
+## ADR Anchors
+
+- **ADR-058** (Brainstorm Gate and Two-Pass Review): the mandatory inspection-before-refund
+  rule is the operational form of two-pass discipline applied to physical goods. The inspection
+  step is pass one (observe without deciding); the disposition step is pass two (decide with
+  full information). Associates who skip inspection and process on customer assertion collapse
+  both passes, producing the same class of error as reviewers who evaluate and generate
+  simultaneously.