Marketplace Risk Model Calibration Review

Use this skill to convert marketplace risk model, fraud score, trust score, abuse classifier, quality risk, payout risk, listing risk, queue threshold, calibration, drift, appeal, and policy feedback questions into a concrete artifact with owners, gates, metrics, and recovery paths.

Workflow

1. Identify risk actions, model outputs, label sources, ground truth windows, reviewer queues, threshold decisions, protected/segment fairness, appeals, drift signals, loss functions, and business constraints.
2. Read references/marketplace-risk-model-calibration-patterns.md.
3. Classify the situation as new model launch, threshold retune, drift investigation, false-positive spike, false-negative loss spike, new abuse pattern, policy change, or appeals quality review.
4. Define calibration dataset, threshold policy, queue routing, human review, appeal loop, fairness checks, drift monitoring, policy feedback, and safe rollout plan.
5. Produce risk model calibration review, state machine, decision table, event schema, threshold matrix, monitoring checklist, and policy-feedback plan.

When not to use

• Do not use for generic advice the base model already handles without this skill's specific artifact contract.

Guardrails