Imported from Prism-Shadow/GDPevo (experiments/claude_code_glm_5_2_max/reports/skills/task_group_002/reflect-3/reflect-3_attempt_01/SKILL.md). Install upstream with npx skills add Prism-Shadow/GDPevo --skill reflect-3_attempt_01. Copyright stays with the author.

MedBridge Sales Ops — CRM Reconciliation & Quote Decision Skill

Transferable workflow rules for solving MedBridge Sales Ops CRM tasks against the remote business API. Covers the two task families: QUOTE (reconcile a customer RFQ/quote into an account-ready quote decision package) and RECONCILIATION (reconcile a won opportunity into a finance-ready engagement reconciliation).

These are reusable SOPs and pitfalls — not answers. Always derive every value from the live API records for the task's specific quote/RFQ/opportunity.

0. Environment & API usage

• Business records live ONLY behind the remote HTTP JSON API (API_BASE_URL). Do not invent records; do not read any local env/ source. Reconcile the task narrative against the API and trust the API over narrative wording when they conflict (e.g., a customer name in the prompt that differs from customers[].name — use the API value; use the API customer_id linked to the quote/RFQ/opportunity even if the prompt names the customer differently).
• Endpoints (all GET unless noted):
  • GET /api — service metadata (collections list).
  • GET /api/<collection> — list all records ({collection, count, records}).
  • GET /api/<collection>?<key>=<value> — filter (case-insensitive, matches nested keys; multiple filters AND). Use this for collections that have no detail-by-id.
  • GET /api/<collection>/<id> — single record by id. Detail-by-id is only available for: customers, products, rfqs, quotes, freight-quotes, opportunities. For invoices, payments, revenue-journals, events, vouchers, policies use the listing/filter endpoints.
  • GET /api/search?q=<text> — substring search across all collections.
• Collections: customers, products, rfqs, quotes, freight-quotes, policies, opportunities, invoices, payments, revenue-journals, events, vouchers.
• Revenue-journals have NO customer_id and will not filter by customer_id — always filter revenue-journals by opportunity_id. (Same safe pattern for invoices/payments/events/vouchers: filter by opportunity_id.)
• Money = USD with 2 decimals. Dates = ISO YYYY-MM-DD. Use stable record IDs exactly as they appear in the API.
• Output: return ONLY valid JSON matching the task's answer_template.json exactly — same field names, same nesting, the controlled enum values declared in the template, no markdown fences, no prose.

1. QUOTE tasks — quote decision package

Trigger: prompt references a quote id (Q-...) or RFQ id (RFQ-...) and asks for revised EXW pricing, freight options/totals, route risk, recommended mode, freight-validity/reconfirmation, and payment terms.

1.1 Record gathering

1. Fetch the quote (/api/quotes/<id>) or RFQ (/api/rfqs/<id>) — note customer_id, quote_date, primary_product_code/requested_modules, confirmed_quantity, incoterm, quote_type.
2. Fetch the customer (/api/customers/<customer_id>) — note segment, customer_type, is_recurring, payment_profile, grant_terms.
3. Fetch the product (/api/products/<code>) — note price_tiers, shelf_life_months, article_number, cold_chain_required.
4. Fetch freight quotes: /api/freight-quotes?quote_id=<quote_id>.
5. Fetch policies: /api/policies (payment_terms, quote_scope, freight, quote_validity, incoterms).

1.2 Catalog tier by quantity (EXW pricing)

• Pick the price_tiers entry where min_qty <= confirmed_quantity AND (max_qty is null OR confirmed_quantity <= max_qty). This single tier gives unit_price_usd, lead_time_days, shelf_life_months.
• exw_total_usd (or line_total / grand_total for EXW-only) = confirmed_quantity × tier.unit_price_usd.
• Do NOT use a "prior_unit_price_usd" from the quote line item — the catalog tier overrides it (revision quotes raise quantity into a cheaper tier).

1.3 Freight options

• Distractor exclusion: drop any freight quote whose id starts with FR-DIS- OR whose destination contains "Distractor". These are never part of the answer.
• Include every REAL freight quote for the quote_id, one per mode (typically AIR, SEA, ROAD). Include a stale/expired real option with its stale flag set — do NOT silently drop it (the stale flag and a client warning are how staleness is reported).
• Per-option fields:
  • freight_id = the freight quote id.
  • mode = UPPER(RI mode) (AIR/SEA/ROAD).
  • freight_cost_usd = cost_usd.
  • transit_days = the freight quote's transit_days_text string verbatim (e.g. "4-6 days").
  • valid_until = freight quote valid_until.
  • grand_total_usd = exw_total_usd + freight_cost_usd (compute even for stale options).
  • Risk enums (derive from route_risk):
    • QUOTE-template risk_level: low→LOW, medium→MEDIUM, high→HIGH.
    • QUOTE-template risk_flag: low→NONE, medium→MEDIUM_BORDER_RISK (high-risk analog: HIGH_BORDER_RISK).
    • DECISION-template customs_border_risk: LOW/MEDIUM/HIGH.
  • DECISION-template staleness fields:
    • source_is_stale = (freight quote status == "stale").
    • validity_status = VALID when valid_until >= quote_date and status active; EXPIRED when valid_until < quote_date (stale/expired).

1.4 Recommended transport mode (KEY RULE)

• Recommended mode = the CHEAPEST VALID freight option by grand_total_usd, considering only VALID (active, non-stale, non-distractor) options.
• Risk level does NOT determine the recommendation — cost among valid options does. A low-risk air option does NOT beat a cheaper valid sea option. Examples: cheapest valid = SEA when SEA is active and cheapest; AIR only when AIR is the sole valid (or cheapest valid) option.
• Stale and distractor options are never recommended.

1.5 Policy / control flags

• freight_reconfirmation_required = true always (policy: freight rates need reconfirmation at order; valid only through valid_until).
• all_freight_options_valid_on_quote_date (QUOTE template) = true only if every included real option has valid_until >= quote_date (i.e., none expired before the quote date). A stale/expired real option makes this false.
• quote_basis:
  • Freight-options quote (incoterm "EXW plus freight options") → "EXW".
  • Indicative EXW-only quote (RFQ with no destination) → "EXW_ONLY". Use the template's controlled placeholder value, NOT the policy terms_code (EXW_ONLY_EXCLUDE_FREIGHT).
• payment_terms:
  • New NGO client (payment_profile NEW_CLIENT_REVIEW, is_recurring false, segment new_ngo) → PREPAY_100.
  • Recurring NGO / commercial (payment_profile NET_30_AFTER_PO) → NET_30_AFTER_PO.
  • Map from the customer's payment_profile / the matching payment policy terms_code.
• customer_policy (where present) = customer segment (e.g. recurring_ngo).
• offer_validity_days = 30 (policy: catalog quote pricing valid 30 calendar days from quote date).
• freight_excluded = true for indicative quotes with no destination (EXW only); false/omit for freight-options quotes.
• who_documentation_required = true for IEHK-family / WHO-standardized emergency-health-kit modules (codes prefixed IEHK-); false for non-WHO products. (Note: products are "IEHK-style" but still trigger the WHO doc flag.)

1.6 Module-level quote discipline (RFQ module quotes)

• Output exactly ONE line item per entry in the RFQ requested_modules, at module granularity: product_code, article_number, quantity, unit_price (single-tier module price), lead_time_days, shelf_life_months, line_total = quantity × unit_price.
• Do NOT expand components into line items, and ignore component_composition_distractors — these are medical-review noise, not quoteable SKUs. The customer must explicitly request component-level pricing to split a module.
• EXW-only module quote: no freight options block; quote_basis EXW_ONLY, freight_excluded true.

2. RECONCILIATION tasks — engagement/finance reconciliation

Trigger: prompt references an opportunity id (OPP-...) and customer id (CUST-...) and asks to reconcile won-amount, milestone invoices, payments, revenue-recognition coverage, outstanding balance, and CRM/accounting/event follow-ups. Use the prompt's stated current business date as as_of_date and for due-date comparisons (default 2026-06-01).

2.1 Record gathering

1. Opportunity: /api/opportunities/<opp_id> — stage, won_amount_usd, outstanding_amount_usd, contact, phases[] (phase_id, amount_usd, completion_date, invoice_id).
2. Customer: /api/customers/<customer_id> — name, segment, contacts.
3. Invoices: /api/invoices?opportunity_id=<opp_id> — amount_usd, status, paid_amount_usd, outstanding_amount_usd, due_date, phase_id.
4. Payments: /api/payments?opportunity_id=<opp_id> — amount_usd, invoice_id, status.
5. Revenue journals: /api/revenue-journals?opportunity_id=<opp_id> (NOT by customer_id) — invoice_id, phase_id, amount_usd, status.
6. Event: /api/events?opportunity_id=<opp_id> — event_date, status, voucher_code, primary_contact, follow_up_owner.
7. Voucher: /api/vouchers?opportunity_id=<opp_id> — discount_percent, max_redemptions, status, valid_until.

2.2 Milestone identification & ordering

• Each opportunity phase = one milestone, ordered ascending by phase order.
• Where the template declares a milestone_id enum (MS1 | MS2 | MS3), use those enum values IN PHASE ORDER — NOT the API phase_id (MER-P1, etc.) and not the invoice id. Map phase 1→MS1, phase 2→MS2, phase 3→MS3.
• Where the template allows a freeform "stable phase or invoice milestone id", use the API phase_id (e.g. HEL-P1).
• Money per milestone: amount/invoice_total = invoice amount_usd; paid_amount/amount_paid = invoice paid_amount_usd; amount_unpaid = invoice outstanding_amount_usd.

2.3 Invoice / payment state enums

• invoice_state: invoice status paid→PAID, unpaid/open→OPEN, void→VOID, unknown→UNKNOWN.
• payment_state: fully paid→PAID, partially paid→PARTIAL, unpaid→UNPAID, unknown→UNKNOWN. A milestone is PAID only if paid_amount_usd equals amount_usd (and a posted payment exists).

2.4 due_date rule (KEY)

• Paid milestone → due_date: null (the obligation is settled; the invoice due date is no longer relevant).
• Unpaid milestone → due_date = the invoice's due_date (ISO string).

2.5 Revenue-recognition state mapping (KEY — missing-journal detection)

For each milestone, set revenue_recognition_status / recognition_status:

• Paid AND a matching revenue-journal exists (journal for that invoice_id/phase_id, status posted) → RECOGNIZED.
• Paid AND NO matching revenue-journal → MISSING_REVENUE_JOURNAL (decision template) / REQUIRED_MISSING (status template). This is the critical defect to detect: a milestone that is complete and paid but whose revenue was never recognized (deferred revenue not moved to income).
• Unpaid → NOT_REQUIRED_UNPAID (no revenue to recognize until paid).
• Unknown/inconsistent → UNKNOWN (decision template only).

Top-level recognition_status (status template):

• All paid milestones have journals → COMPLETE_FOR_PAID_MILESTONES.
• Any paid milestone lacks a journal → MISSING_FOR_PAID_MILESTONES.
• recognized_milestones = list of milestone ids with journals; missing_required_milestones = paid milestones lacking journals; recognized_amount = sum of recognized journal amounts.

2.6 Balance & match booleans

• total_paid_amount = sum of invoice paid_amount_usd.
• outstanding_balance = won_amount − total_paid_amount (= sum of invoice outstanding_amount_usd). Must agree with opportunity outstanding_amount_usd.
• opportunity_matches_milestones / opportunity_matches_phase_total = (sum of phase amounts == won_amount_usd).

2.7 Follow-up routing (collection vs accounting vs event-invitation)

Accounting action — based on revenue-recognition need:

• A paid milestone MISSING its revenue journal → RECORD_REVENUE_MSx (milestone_id = that MS, amount = its invoice amount, debit_account = DEFERRED_REVENUE, credit_account = IMPLEMENTATION_SERVICES_REVENUE, owner_queue = ACCOUNTING). This is the primary_accounting_action when present.
• Paid milestone already recognized → VERIFY_REVENUE_ONLY (no new posting; owner ACCOUNTING).
• Unpaid milestones → NO_ACCOUNTING_ACTION (owner_queue NONE).

Collection action — based on the unpaid milestone vs current business date:

• Unpaid milestone with due_date <= current business date (due/overdue) → SEND_COLLECTION_NOTICE (owner_queue = COLLECTIONS, due_date = invoice due date, amount = outstanding).
• Unpaid milestone with due_date > current business date (not yet due) → MONITOR_UNPAID_NOT_DUE (owner_queue = ACCOUNT_MANAGEMENT, due_date = invoice due date, amount = outstanding).
• No unpaid milestone → NO_COLLECTION_ACTION (owner_queue NONE).

Event-invitation action — based on the linked event:

• Event status scheduled/confirmed and upcoming (event_date >= current date) with invitation not yet sent → SEND_BRIEFING_INVITE (owner_queue = the event's follow_up_owner, typically ACCOUNT_MANAGEMENT; event_id, voucher_code, contact_name, customer_id populated).
• Invitation already sent / event active → VERIFY_INVITE_SENT.
• No event → NO_INVITE_ACTION.
• Event status enum: scheduled→SCHEDULED, confirmed→ACTIVE/CONFIRMED per template, completed→COMPLETED, cancelled→CANCELLED.
• Voucher: voucher_status from voucher status (active→ACTIVE, draft→ DRAFT, expired→EXPIRED, disabled→DISABLED); discount_amount = voucher discount_percent value; max_uses = voucher max_redemptions.

2.8 Contact linkage

• primary_contact/contact = the opportunity's contact name; link to the customer via customer_id (and opportunity_id where the template field exists). Tie every follow-up task to this contact name and the customer/opportunity ids.

3. Cross-cutting pitfalls (from reflect feedback)

1. Recommended mode is cheapest-VALID, not lowest-risk. Picking the low-risk air option over a cheaper valid sea option is wrong. (Cost among active/non-stale/non-distractor options decides.)
2. Stale real freight options stay in the array, flagged — do not drop them. Only distractors (FR-DIS-* / "Distractor" destination) are removed.
3. Paid milestones get due_date: null. Leaving the invoice due date on a paid milestone is wrong.
4. Detect missing revenue journals explicitly. A paid milestone with no matching revenue-journal is MISSING_REVENUE_JOURNAL and drives a RECORD_REVENUE accounting action — this is the single most important reconciliation finding.
5. Collection routing depends on due date vs current business date. A not-yet-due unpaid milestone is MONITOR_UNPAID_NOT_DUE (account management), not a collection notice.
6. Filter revenue-journals (and other finance records) by opportunity_id, not customer_id. Revenue journals carry no customer_id.
7. quote_basis uses the template's controlled value, not the policy terms_code (e.g. EXW_ONLY, not EXW_ONLY_EXCLUDE_FREIGHT).
8. who_documentation_required is true for IEHK-family modules even though the catalog calls them "IEHK-style".
9. Module RFQs stay at module line level — never expand components or honor component_composition_distractors.
10. Catalog tier overrides prior unit price on revision quotes; recompute EXW total from the tier that matches the confirmed quantity.
11. Trust API values over prompt narrative for customer name, ids, amounts, and statuses; reconcile which entity the prompt asks about, then read facts from that API record.
12. Use detail-by-id only for the six id-keyed collections; otherwise list
    • filter. Use stable API ids verbatim.
13. Freeform string fields (warning messages, task titles) are not reliably matchable — prioritize getting every structured id, amount, date, and enum exactly right; for freeform warning text, state the concrete finding concisely (which freight id is stale/expired, its valid_until vs quote date, the recommendation, the reconfirmation requirement).
14. Output ordering: milestones ascending by milestone_id (MS1, MS2, MS3); freight options in AIR, SEA, ROAD order unless the template implies otherwise.