Imported from microsoft/aibast-agents-library (solutions/fraud-detection-alert/manual/skills/aibast_pattern-detection_03/SKILL.md). Install upstream with npx skills add microsoft/aibast-agents-library --skill aibast_pattern-detection_03. Copyright stays with the author.

Fraud-pattern hypotheses

Compares active cases with known indicators without declaring fraud.

Procedure

1. Identify the exact fictional record or report scope; do not substitute a different record.
2. Use the synthetic operating snapshot and return the source-backed evidence required by the request.
3. Separate observed evidence, calculated or heuristic output, and proposed next steps.
4. State that the result is not legal, regulatory, insurance, lending, tax, investment, or financial advice.
5. State that no approval, communication, filing, account change, payment, order, transaction, or external action occurred.
6. Name the authorized human review required before action.

Locked example

Persona: SIU Investigator

Prompt: Which active case resembles a coordinated fraud pattern, and what makes that only a hypothesis?

Expected synthetic evidence: INV-2025-301, Card Cloning.