Build the complete application

## User
A fraud finance analyst reconciling operating results.

## Build
Reconcile synthetic fraud losses, reversals, recoveries, and review costs. Keep observed accounting outcomes separate from claims that a model prevented a loss.

## Input Contract
- Key: economic_event_id + event_version.
- Fields: transaction_id, gross_loss, recovery, reversal, cost, recognized_at, currency, link_to_original.
- Model-estimated prevention is a separate nonaccounting field. Do not net duplicated recovery events twice.

## Workflow
1. Match lifecycle events to the original economic exposure.
2. Reconcile gross loss, eligible recoveries, and operating cost.
3. Identify duplicated, unmatched, and censored events.
4. Export a finance bridge and unresolved accounting items.

## Interface
A loss-to-net-cost waterfall with event lineage, unmatched records, and realized-versus-estimated panels.

## Deliverables
- An idempotent fraud loss event ledger.
- A finance reconciliation workbook with explicit attribution limitations.
- Runnable local application, setup commands, and a dependency lockfile.
- Versioned input data, data dictionary, acquisition instructions, and rights notes.
- Calculation traces, source references, assumption records, and exported review artifacts.
- A deterministic baseline and tests for normal, incomplete, contradictory, and failure-recovery cases.
- A recorded complete workflow with a reproducible manifest and explicit limitations.

## Acceptance Criteria
- Repeated event ingestion cannot reduce losses twice.
- Recoveries require a linked original exposure and currency.
- A declined transaction is not automatically a prevented fraud loss.
- Observed net loss and modeled avoided loss are reported separately.

## Contrast Case
The same gross loss produces different observed net cost after verified recovery. This does not prove model effectiveness.

## Dataset Scope
The attached starter files are prototype inputs. Expected starter answers are visible. They are not independent holdouts.

Full target: Author fictional loss, recovery, fee, and reversal events with stable identifiers. Public synthetic fraud labels do not contain the full accounting lifecycle.

Complete target contracts and expert review remain pending. Public observations are not historical release vintages.

## Tools
DuckDB, pandas, openpyxl

Use an available model appropriate to the task. Record provider, model identifier, version or retrieval date, settings, cost, and latency. Use deterministic code for calculations. Models may extract, retrieve, or explain evidence. They must not invent missing records, override approval rules, or claim professional authority.

## Constraints
Accounting reconciliation alone cannot identify the causal effect of a fraud intervention.

- No fabricated records, source citations, professional endorsements, or expert-adjudication claims.
- No future-data leakage into dated decisions, model fitting, or historical valuation inputs.
- No unauthorized payments, account restrictions, regulatory filings, or production changes.
- No automated appraisal certification, tax-compliance determination, or unsupported fraud accusation.
- No concealed uncertainty or silently invented inputs required for a numerical conclusion.

## Scoring
- Correct domain calculations and justified methods with documented assumptions. Weight: 30.
- Source lineage, temporal integrity, rights checks, and explicit synthetic-data labels. Weight: 25.
- A working application with review, export, error handling, and reproducible state. Weight: 20.
- Adversarial cases, uncertainty, missing-data behavior, and sensitivity analysis. Weight: 15.
- Evidence-linked conclusions that distinguish facts, assumptions, and unresolved judgments. Weight: 10.
Score each dimension from zero to two. Weighted score = sum(weight * score / 2). A failed safety gate blocks acceptance.