Support Triage Policy Scorer

Scenario

Use after intake normalization when a support team needs predictable category, urgency, risk, and service target fields.

Input and output

Input contains the normalized contract and a redacted policy with keyword groups, risk terms, and SLA hours. Output has category, urgency, sentiment hint, risk flags, SLA target, and score_hash.

Execution and recovery

Run python scripts/score.py examples/input.json. Missing policy groups or invalid SLA values return visible errors. Change the buyer policy and rerun; this deterministic scorer does not call a model.

Privacy boundary and acceptance

Do not include customer content beyond a redacted test string. The sample must identify a billing-like category, high urgency, a risk flag, and an SLA target.