Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
invoice date, due date, posting date, order date, expected arrival, done date
Choose the date based on the requested analysis, not column name similarity alone.
Workflow
Read the specific Odoo TM/QM before choosing state, payment_state, invoice_status, stage, or type.
Read the domain map before adding Odoo route fixtures; keep accounting/payment and sales/purchase in separate fixture packs.
If a request combines lifecycle and settlement wording, require both fields when both are available.
If a metric depends on accounting policy, receivable aging bucket, risk rating, or budget/version definition, route to clarification until modeled.
Add AI route fixtures for one Odoo subdomain at a time; current packs cover accounting/payment, sales/purchase, stock picking, and MRP.
Use foggy-model-generate for model changes and keep Odoo-specific business rules in this domain layer.
Use this bridge repo as the local debug/evidence workspace for Odoo TM/QM changes. After a TM/QM is verified, promote the model artifacts to https://github.com/foggy-projects/foggy-model-registry.git; do not treat foggy-dataset-demo/src/main/resources/foggy/templates/odoo as the long-term source of truth.
Current Evidence
Sales/purchase, stock picking, MRP, accounting/payment, and first negative semantic-conflict direct fixtures have local passing evidence.
mrp_production and mrp_bom are available in the local fixture database for MRP direct tests.
For manufacturing completion throughput, use state=done plus dateFinished$yearMonth; do not use start/deadline/create/write dates as completion-period substitutes.
For customer invoice amount/count trends, use state=posted, moveType=out_invoice, and invoiceDate$yearMonth; do not use due date, posting date, create date, or write date as invoice-period substitutes.
For customer credit notes / refunds, use moveType=out_refund; do not include them in ordinary customer invoice trend unless explicitly requested.
For vendor credit notes / refunds, use moveType=in_refund; do not treat them as ordinary vendor bills (in_invoice).
For reversed/refund settlement, use paymentState=reversed while keeping accounting lifecycle state=posted; do not use state=reversed.
For paid invoices, require paymentState=paid and keep state=posted as accounting lifecycle; do not use state=paid.
For paid vendor bills, require moveType=in_invoice, state=posted, and paymentState=paid; do not use customer invoice direction.
For open vendor bills, require moveType=in_invoice, state=posted, and paymentState in ('not_paid', 'partial', 'in_payment'); do not include draft bills or customer invoices.
For overdue payables, use apOverdueAmount and apOverdueVendorCount; do not reuse AR metrics or invoice-date period logic.
For vendor payments, require paymentType=outbound and partnerType=supplier; use isReconciled for matched/settled payment state, not generic document state.
For "which payment matched which vendor bill" questions, use OdooAccountPaymentBillMatchQueryModel; do not rely on isReconciled=true alone as proof of a specific bill match.
For matched vendor payments where the bill remains open, keep the payment-bill match fact and require ; preserve bill residual amount instead of treating the matched payment as fully paid.
Do Not
Do not infer Odoo enum meaning from English words alone; respect dictRef and field descriptions.
Do not use invoice due date as business date for revenue trend unless the model contract requests aging/overdue analysis.
Do not hard-code Odoo field names into engine validation.
billMove$paymentState=partial
For split vendor payments, use one payment-bill match row per bill; the same payment$caption can return multiple vendor bill rows.
For purchase-order to receipt to vendor-bill questions, use OdooPurchaseDocumentFlowQueryModel; do not infer the cross-document relationship only by string-matching origin / invoice_origin.
For received-but-open payable purchase flows, require purchase lifecycle (purchaseOrder$state in ('purchase','done')), receipt completion (receiptPicking$state=done), vendor bill direction (billMove$moveType=in_invoice), and bill settlement (billMove$paymentState in ('not_paid','partial','in_payment')).
For sales-order to delivery to customer-invoice to customer-payment questions, use OdooSaleDocumentFlowQueryModel; do not infer the cross-document relationship only by string-matching origin / invoice_origin.
For delivered-but-not-fully-collected sales flows, require sales lifecycle (saleOrder$state in ('sale','done')), delivery completion (deliveryPicking$state=done), customer invoice direction (invoiceMove$moveType=out_invoice), invoice settlement (invoiceMove$paymentState in ('partial','in_payment') when customer payment evidence is requested), and inbound customer payment (payment$paymentType=inbound, payment$partnerType=customer).
For overdue AR customer count, use COUNT_DISTINCT(if(..., null)) in QM formulas; do not use SQL-style count(distinct(...)) inside the expression DSL.
