# Reflect 3 Attempt 01 > SKILL: MedBridge Sales Ops — CRM/B2B Quote & Engagement Reconciliation - Skill: `prism-shadow/reflect-3-attempt-01-21` (Agent Skill) - Install (CLI): `npx skillmds@latest add prism-shadow/reflect-3-attempt-01-21` - Raw SKILL.md: https://api.skillmd.com/api/skills/prism-shadow/reflect-3-attempt-01-21/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: prism-shadow (https://skillmd.com/u/prism-shadow) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/prism-shadow/reflect-3-attempt-01-21 --- # SKILL: MedBridge Sales Ops — CRM/B2B Quote & Engagement Reconciliation ## Description / when this applies Use this skill when you are handed ONE task in the **MedBridge Sales Ops** domain: a CRM / B2B medical-supply account that needs either (a) an **EXW quote decision package** (catalog pricing + freight comparison) or (b) an **engagement / milestone reconciliation** (opportunity vs invoices vs payments vs revenue journals + event/voucher follow-ups). You get a prompt naming specific record IDs, the read-only data API, and an `answer_template.json`. Your job: read the real records from the API, apply the business rules below, and return ONLY JSON that fills the template exactly. Return ONLY the JSON object that matches the provided `answer_template.json`. No markdown, no prose, no code fences, no trailing commentary. Fill every field; never leave template placeholders (`""`, `0`, `<...>`, `YYYY-MM-DD`). --- ## The data API - Base URL is provided by the runner (env var named e.g. `API_BASE_URL` / `BASE_URL` / `http://127.0.0.1:`). Read-only JSON over HTTP (use curl or python urllib). - Collections (GET): `/api/customers`, `/api/products`, `/api/rfqs`, `/api/quotes`, `/api/freight-quotes`, `/api/policies`, `/api/opportunities`, `/api/invoices`, `/api/payments`, `/api/revenue-journals`, `/api/events`, `/api/vouchers`. - Single record: `/api//` (e.g. `/api/quotes/Q-...`, `/api/products/`, `/api/opportunities/OPP-...`). - Filtering: list endpoints accept `?field=value` (nested keys are flattened), e.g. `/api/invoices?opportunity_id=OPP-X`, `/api/freight-quotes?quote_id=Q-X`, `/api/vouchers?code=ABC`. Matching is case-insensitive; numbers match their 2-decimal form. - `/api/search?q=` does full-text search across all collections — use it to resolve a name to an ID (e.g. search the customer name to get the `CUST-...` id). ### Which endpoint answers which question - Customer record / payment profile / type / recurring flag → `/api/customers/`. - Catalog pricing tiers, lead time, shelf life, article_number, components → `/api/products/`. - Confirmed quantity, quote date, primary product, destination → `/api/quotes/`. - Requested modules + quantities, indicative flag, destination → `/api/rfqs/`. - Transport options → `/api/freight-quotes?quote_id=` (NOT by customer_id — see gotcha). - Business rules / payment terms / validity → `/api/policies` (one list, read all 8 rules). - Opportunity stage, won amount, phases, outstanding → `/api/opportunities/`. - Milestone invoices → `/api/invoices?opportunity_id=` (each has phase_id, due_date, status, paid_amount). - Cash received → `/api/payments?opportunity_id=`. - Revenue recognition coverage → `/api/revenue-journals?opportunity_id=`. - Celebration/briefing event → `/api/events?id=` or `?customer_id=`. - Voucher / offer controls → `/api/vouchers?code=`. ### Gotchas resolving data - **Freight is keyed by `quote_id`, not `customer_id`.** `?customer_id=...` returns 0 freight rows. - Resolve IDs from the prompt: the prompt usually gives the quote/rfq/opportunity id, the customer id, contact name, event id, and voucher code directly. If only a name is given, use `/api/search?q=`. - Read money/dates straight from records; do not recompute unit prices — pick the catalog tier. --- ## Output conventions (confirmed) - **Money**: numbers with 2-decimal precision (e.g. `42480.00`, `76.0`). USD. Totals must reconcile exactly (no rounding drift). - **Dates**: ISO `YYYY-MM-DD`. - **Enums**: use the EXACT controlled values printed in the template (usually UPPERCASE like `WON`, `PAID`, `OPEN`, `RECOGNIZED`, `SEND_BRIEFING_INVITE`). Do not invent synonyms. - **Booleans**: real `true`/`false`. - **null**: use JSON `null` where the template says "or null" and the value is not applicable (see paid-milestone due_date rule). - A `*_policy` / policy-reference field expects the **policy record ID** (e.g. `POL-RECURRING-NGO-PAYMENT`), not prose. - Copy IDs/codes verbatim from the records (case-sensitive: `CUST-...`, `OPP-...`, `FR-...`, `MS1`, voucher codes). --- ## FAMILY A — EXW Quote Decision Package (catalog pricing + freight) Recognize it by: a `Q-...` quote (or `RFQ-...`) about units of a product, asking for EXW pricing, freight options, risk, recommended mode, validity, payment terms. ### SOP 1. GET the quote (`/api/quotes/`): read `customer_id`, `quote_date`, `primary_product_code`, and `confirmed_quantity`. 2. GET the product (`/api/products/`). Pick the **price tier whose [min_qty, max_qty] contains the confirmed quantity** (max_qty `null` = open-ended top tier). That tier gives `unit_price_usd`, `lead_time_days`. `shelf_life_months` is product-level. Ignore the prior quote's unit price — the current catalog tier overrides it. 3. `exw_total = unit_price * confirmed_quantity` (2 decimals). 4. GET freight `/api/freight-quotes?quote_id=`. Keep only the **real options for this quote**: typically one each of mode air/sea/road with `status:"active"` and a real destination. **EXCLUDE** any record whose id contains `FR-DIS-` or whose `destination` is `"Distractor route"`, and any whose `status` is `stale`/`mismatch` *when a distractor*. (A genuine quote option that is itself stale/expired is still listed — see step 6.) 5. For each kept freight option: - `freight_cost_usd` = `cost_usd`; `transit_days` = the `transit_days_text` (e.g. "4-6 days"); `valid_until` = record's `valid_until`. - `grand_total = exw_total + freight_cost` (2 decimals). - Risk from `route_risk`: `low→LOW`, `medium→MEDIUM`, `high→HIGH`. If a template uses a `risk_flag`/`customs_border_risk`, map: low→`NONE`, medium→`MEDIUM_BORDER_RISK`, high→border/high risk (match the template's enum exactly). 6. **Validity / staleness**: an option is valid on the quote date iff `valid_until >= quote_date` AND `status == "active"`. A `status:"stale"` / expired (`valid_until < quote_date`) genuine road option must be **flagged** (set its validity/stale flags, e.g. `source_is_stale: true`, validity status STALE/EXPIRED, and the top-level `road_quote_invalid_or_stale: true` + a `freight_warning` string). Set `all_freight_options_valid_on_quote_date` accordingly. 7. **Recommended mode** (rewarded rule): choose the option with the **lowest grand_total among options that are valid (not expired/stale) AND `route_risk == low`**. Tie-break by lowest cost. (Confirmed: it is NOT "fastest" and NOT simply the global cheapest — a cheaper medium-risk or stale option does NOT win.) - Example logic: if AIR and SEA are both low-risk, recommend the cheaper (SEA); if only AIR is low-risk (SEA medium, ROAD stale/high), recommend AIR. 8. **Freight reconfirmation**: always `true` for catalog freight (policy `POL-FREIGHT-RECONFIRM`, "reconfirm at order; valid only through valid_until"). 9. **Payment terms** by account (from customer record + policies): - Recurring NGO / recurring commercial with an after-PO net profile → `NET_30_AFTER_PO` (policy `POL-RECURRING-NGO-PAYMENT`). - New / prospect NGO (not recurring, segment new_*) → `PREPAY_100` (policy `POL-NEW-CLIENT-PAYMENT`). - When in doubt, the customer's `payment_profile` field is the source of truth; map it to the matching policy `terms_code`. 10. `quote_basis` is `EXW` (or `EXW_ONLY` per template) for these packages. ### FAMILY A-modules — Indicative module RFQ (EXW only, freight excluded) Recognize it by: an `RFQ-...` for several module product codes, "indicative", "destination pending/no destination", "EXW only", "module level". - Quote at **module line level only**. The RFQ may include `component_composition_distractors` / composition tables — **ignore them; do NOT split modules into component SKUs** (policy `POL-MODULE-GRANULARITY`). - One line per requested module: `product_code`, `article_number` (from product, keep as the string the API returns, e.g. "500101"), `quantity` (from RFQ), `unit_price` (single tier), `lead_time_days`, `shelf_life_months`, `line_total = unit_price * quantity`. - `grand_total` = sum of line_totals. - No destination ⇒ `quote_basis = EXW_ONLY`, `freight_excluded = true` (policy `POL-INDICATIVE-EXW`). Do not fetch/attach freight. - `payment_terms = PREPAY_100` for a new NGO account. - `offer_validity_days = 30` (policy `POL-QUOTE-VALIDITY`: catalog pricing valid 30 days). - `who_documentation_required = true` for these WHO/IEHK-style emergency health kits. --- ## FAMILY B — Engagement / Milestone Reconciliation Recognize it by: an `OPP-...` opportunity + `CUST-...`, asking to reconcile won amount vs milestones, summarize invoice/payment/outstanding/revenue-recognition state, route CRM/ accounting follow-ups, and include a related event + voucher. ### SOP 1. GET opportunity `/api/opportunities/`: `stage` (`closed_won → WON`), `won_amount_usd`, `phases[]` (each has phase_id, amount_usd, completion_date, invoice_id), `outstanding_amount_usd`, `contact`. 2. GET customer for `customer_name` and the contact's linkage. 3. `phase_total = sum(phase amounts)`. `opportunity_matches_phase_total = (phase_total == won_amount)`. 4. GET invoices `/api/invoices?opportunity_id=`, payments `/api/payments?opportunity_id=`, revenue-journals `/api/revenue-journals?opportunity_id=`. Join by `phase_id` / `invoice_id`. 5. `total_paid = sum(invoice.paid_amount_usd)`; `outstanding_balance = sum(invoice.outstanding_amount_usd)` (should equal the opportunity's outstanding). 6. Build one milestone per phase, **ordered ascending** (MS1, MS2, MS3 …). If the template's `milestone_id` is an enum (`MS1|MS2|MS3`), use that ordinal id — NOT the phase_id. If the template instead asks for a "stable phase/invoice milestone id", use the phase_id. Per milestone: - `amount` = phase amount; `paid_amount` = invoice paid_amount; `amount_unpaid` = invoice outstanding. - `invoice_state`: invoice `status` paid→`PAID`, unpaid→`OPEN` (void→`VOID`). - `payment_state`: fully paid→`PAID`, none→`UNPAID`, between→`PARTIAL`. - **`due_date`: for a PAID milestone set `due_date = null`; for an UNPAID milestone use the invoice `due_date`.** (Confirmed convention.) - `recognition_status`: * paid AND a revenue-journal exists for that phase → `RECOGNIZED`. * paid AND **no** revenue-journal → `MISSING_REVENUE_JOURNAL` (also `REQUIRED_MISSING` style in other templates). * unpaid → `NOT_REQUIRED_UNPAID`. 7. **Revenue recognition rollup** (policy `POL-REVREC`: recognize only paid+complete milestones): - `recognized_milestones` = paid milestones that have a journal; `recognized_amount` = their sum. - `missing_required_milestones` = paid milestones with NO journal. - `recognition_status`: all paid milestones recognized → `COMPLETE_FOR_PAID_MILESTONES`; some paid milestone missing a journal → `MISSING_FOR_PAID_MILESTONES`; nothing paid → `NOT_REQUIRED`. 8. **Accounting action routing** (separate from collections): - If a **paid milestone is missing its journal** → `primary_accounting_action = RECORD_REVENUE_MS` for that milestone. The `accounting_action`: `action = RECORD_REVENUE_MS`, `milestone_id = MS`, `amount` = that phase amount, `debit_account = DEFERRED_REVENUE`, `credit_account = IMPLEMENTATION_SERVICES_REVENUE`, `owner_queue = ACCOUNTING`. (Mirror the revenue-journal pattern: debit Deferred Revenue, credit Implementation Services Revenue.) - If all paid milestones already recognized → `VERIFY_REVENUE_ONLY` / `NO_ACCOUNTING_ACTION` with `milestone_id = NONE`, `amount = 0.00`, accounts `NONE`, `owner_queue NONE` (per the template's enums). 9. **Collection action routing** (drive only from UNPAID milestones; compare due_date to the current business date `as_of`): - Unpaid and **not yet due** (`due_date > as_of`) → `collection_action = MONITOR_UNPAID_NOT_DUE`; `collection_task`: `action = MONITOR_UNPAID_NOT_DUE`, `milestone_id = MS`, `amount` = unpaid amount, `due_date` = invoice due, **`owner_queue = ACCOUNT_MANAGEMENT`** (NOT COLLECTIONS), `contact_name` = the contact. - Unpaid and **due/overdue** (`due_date <= as_of`) → `SEND_COLLECTION_NOTICE`, and `owner_queue = COLLECTIONS`. - No unpaid milestones → `NO_COLLECTION_ACTION`. 10. **Event + voucher + invite** (the briefing/celebration): - `event_status` from event `status` (e.g. scheduled→`SCHEDULED`, confirmed→`SCHEDULED`/per template enum; active→`ACTIVE`). - voucher: `voucher_status` from status (active→`ACTIVE`), `discount_amount = voucher.discount_percent`, `max_uses = voucher.max_redemptions`. - `invite_action = SEND_BRIEFING_INVITE` (or `SEND_EVENT_INVITATION` per template); `invite_task` `owner_queue = ACCOUNT_MANAGEMENT` (event `follow_up_owner` = "Account Management"), with `event_id`, `voucher_code`, `contact_name`, `customer_id` filled. 11. **Follow-up tasks** (when the template uses a single `follow_up_tasks[]` list rather than split action objects): emit one `COLLECTION` task per unpaid milestone (`next_action = COLLECT_UNPAID_MILESTONE`, `milestone_id` set, `amount_due` set, `event_id`/`voucher_code` = null, `due_date` = invoice due) and one `EVENT_INVITATION` task (`next_action = SEND_EVENT_INVITATION`, `milestone_id`/`amount_due` = null, `event_id`/`voucher_code` set, `due_date` = event_date). Always set `linked_customer_id`, `linked_opportunity_id`, `contact_name` on every task. --- ## Business-rule cheat sheet (what the grader rewarded) - **Tier selection**: the tier bracket that contains the confirmed quantity; catalog tier overrides any prior/quoted unit price. - **EXW & grand-total math**: `exw = unit_price * qty`; `grand_total = exw + freight_cost`; 2 decimals; totals reconcile. - **Risk level vs risk flag** are distinct: a numeric/ordinal level (LOW/MEDIUM/HIGH) plus a named flag (NONE / MEDIUM_BORDER_RISK). Set both from `route_risk`. - **Recommended mode = lowest grand_total among VALID, LOW-route-risk options** (feasibility filters first, cost decides among the survivors). Not fastest, not cheapest-overall. - **Feasibility vs risk separation**: exclude stale/expired and distractor freight from the recommendation, but still LIST genuine stale options with their warning flags. - **Validity / reconfirmation**: catalog pricing valid 30 days; freight valid only through its `valid_until`; freight reconfirmation always required at order. - **Payment terms by account type**: recurring → NET_30_AFTER_PO; new/prospect → PREPAY_100; fall back to the customer's `payment_profile`. - **Revenue recognition**: recognize ONLY paid milestones; a **paid milestone missing its journal triggers RECORD_REVENUE** (debit Deferred Revenue → credit Implementation Services Revenue, ACCOUNTING queue); unpaid milestones are never recognized. - **Collection vs accounting are routed separately**: missing-journal = accounting; unpaid = collection. Unpaid-not-due → MONITOR (ACCOUNT_MANAGEMENT); unpaid-overdue → SEND_COLLECTION_ NOTICE (COLLECTIONS). - **Paid-milestone due_date = null**; unpaid milestone keeps invoice due_date. - **Module-only granularity**: never expand modules into components. - **Freight excluded when no destination**: indicative EXW_ONLY quotes carry no freight. ## Common misjudgments / exclusions - Do NOT include `FR-DIS-*` / "Distractor route" freight, "wrong shipment size"/"old route"/ "prior count" benchmark rows, or freight belonging to a different `quote_id`. - Do NOT carry a stale/expired freight rate into the recommendation; only into the warning. - Do NOT over-expand RFQ modules into component SKUs (composition tables are distractors). - Do NOT filter freight by `customer_id` (returns nothing); filter by `quote_id`. - Do NOT reuse the prior/old unit price; always re-pick the current catalog tier. - Do NOT set a paid milestone's due_date to the invoice date — use null. - Do NOT route a not-yet-due unpaid milestone to COLLECTIONS — that is ACCOUNT_MANAGEMENT monitoring until the due date passes. - Match enum spelling/case to the template exactly; a near-synonym is graded wrong. ## Final reminder Return ONLY the JSON object matching the provided `answer_template.json`, every field filled with real values from the API and the rules above. No markdown, no explanation.