Plan Fraud Decision Intelligence workbench as downstream runtime consumer

[mdheller]

Purpose

Plan a Prophet Platform Fraud Decision Intelligence workbench that consumes upstream doctrine and contracts rather than inventing fraud semantics locally.

Upstream doctrine anchor:

• SocioProphet/ProCybernetica:docs/capabilities/fraud-decision-intelligence-control-plane.md

Workbench surfaces

Candidate runtime surfaces:

• fraud event view;
• entity and relationship view;
• evidence bundle view;
• model-vs-rule comparison;
• residual-fraud lift dashboard;
• scenario registry;
• false-positive and false-negative atlas;
• threshold and review-capacity simulator;
• drift and displacement monitor;
• analyst review queue;
• decision receipt viewer.

Required dependency boundaries

Prophet Platform should consume:

• ProCybernetica doctrine and claim boundaries;
• Ontogenesis fraud ontology once available;
• Semantic SerDes fraud envelopes once available;
• model-governance-ledger fraud records once available;
• agent-registry fraud admission profile once available.

Non-goals

This issue does not authorize:

• live fraud scoring;
• production customer decisions;
• denial/suspension actions;
• ingestion of private customer evidence into public fixtures;
• local invention of ontology terms that belong upstream.

Acceptance criteria

• Add a platform design note for the Fraud Decision Intelligence workbench.
• Document all upstream contract dependencies.
• Define UI surfaces as read-only/report-only until policy and runtime gates exist.
• Add negative implementation note: model score cannot be displayed as proof of fraud.
• Add negative implementation note: graph path cannot be displayed as attribution or intent proof.