Northstar Payer Operations
Shared payer-operations environment for Northstar Health Plan. The environment is read-only and provides business APIs plus a SQL query endpoint. All tasks consume environment records and return structured JSON.
Quick start
Every task shares the same environment access pattern. Use the base URL and credentials provided in the task prompt or task_context.json payload:
Base URL: <TASK_ENV_BASE_URL>
SQL endpoint: POST {base_url}/sql/query
Auth header: Authorization: Bearer pa-review-token-014
Primary access method: SQL. Use POST /sql/query for most data retrieval. The business REST endpoints are available for direct lookups by ID. See schema.md for the full 19-table schema.
Run SQL queries with the bundled script:
python3 skill/scripts/query.py "<SQL>" --base-url <BASE_URL> --token pa-review-token-014
Or directly with curl:
curl -s -X POST <BASE_URL>/sql/query \
-H "Authorization: Bearer pa-review-token-014" \
-H "Content-Type: application/json" \
-d '{"sql": "<SQL>"}'
Core workflow
Discover the target: Identify the case, claim, appeal, or queue ID from the prompt. Cross-reference against task_context.json when present.
Query the environment: Pull all relevant records. Start broad (the target record joined with its dependents), then narrow with WHERE clauses.
Evaluate against criteria: When a policy applies, match policy_criteria against case_criteria results and supporting document_facts. Check is_current on documents to separate active evidence from stale.
Apply the source-precedence rule: Every task has a source_precedence rule that controls how conflicting or overlapping records are resolved. See precedence.md.
Build the answer: Fill the template fields from answer_template.json. Currency values rounded to 2 decimal places. Ratios to 4 decimal places. Null for absent modifiers, not empty string. ISO 8601 dates. Lists ordered as specified by the template.
Construct the basis audit: Every answer ends with a basis_audit containing the source-precedence rule, the ordered record precedence trail, the controlling record IDs, and exception/gap record IDs. See precedence.md.
Task domains
The environment supports five task domains, each with specific data models and evaluation patterns. Detailed workflows are in workflows.md.
| Domain |
Typical Entry |
Key Tables |
Precedence Rule |
| UM nurse prior-auth |
case_id |
cases, request_lines, documents, document_facts, policy_criteria, case_criteria, members, plans, authorizations |
current_clinical_records_over_stale_export |
| Pharmacy appeals |
appeal_id |
appeals, cases, drug_trials, assistance_screen, documents, policy_criteria |
payer_appeal_before_manufacturer_assistance |
| Payment integrity |
claim_id |
claims, claim_lines, payment_benchmarks, plans |
effective_benchmark_by_plan_modifier_and_date |
| Peer-to-peer |
p2p_event_id |
p2p_events, cases, request_lines, documents, policy_criteria |
new_patient_specific_p2p_information |
| Margin queue |
queue_id |
service_margin |
margin_threshold_then_charge_sensitivity |
Policies
Five active policies cover the task domains. See policies.md for policy-to-criterion mappings.
| Policy ID |
Name |
Domains |
| POL-PT-LUMBAR-2026 |
Lumbar Physical Therapy Medical Necessity |
UM nurse prior-auth |
| POL-DRUG-EXC-2026 |
Specialty Drug Coverage Exception Appeal |
Pharmacy appeals |
| POL-PET-MPI-2026 |
PET Myocardial Perfusion Imaging Medical Necessity |
P2P |
| POL-CLAIM-RATE-2026 |
Outpatient Imaging and Surgery Payment Benchmark |
Payment integrity |
| POL-ST-PEDS-2026 |
Pediatric Speech Therapy Prior Authorization |
(not used in train tasks) |
Numeric conventions
- Currency (USD): JSON number rounded to 2 decimal places.
- Ratios (e.g. revenue_to_cost_ratio): JSON number rounded to 4 decimal places.
- Units (e.g. approved_units, visits): integer.
- Null modifier: Use
null, not empty string.
- Dates: ISO 8601
YYYY-MM-DD. Months as YYYY-MM.
Reference files
- schema.md — Full database schema with all 19 tables, columns, types, and keys.
- precedence.md — Source-precedence rules and the basis_audit construction pattern.
- policies.md — Policy framework with criteria-key to criterion-ID mappings.
- workflows.md — Step-by-step workflows for each of the five task domains.
1---2name: northstar-payer-ops-263description: "Northstar Health Plan payer operations environment for utilization management, appeals, payment integrity, peer-to-peer review, and margin analysis. Use this skill when working with the Northstar shared payer-operations API at the task-provided base URL. Covers: (1) UM nurse prior-authorization determinations with criteria review, document classification, and authorization issuance, (2) pharmacy coverage-exception appeals with drug-trial evaluation, packet-gap analysis, and manufacturer-assistance screening, (3) payment integrity claim repricing against current benchmarks with line-level correction, (4) peer-to-peer coordinator summaries after completed P2P discussions including adverse-determination deadlines, and (5) UM-finance therapy margin queue analysis with threshold and charge-sensitivity flagging. Every task produces a structured JSON answer with a basis_audit trail."4license: MIT5compatibility: designed for deepagents-code6metadata:7 source: https://skillmd.com/skills/prism-shadow/northstar-payer-ops-268---910# Northstar Payer Operations1112Shared payer-operations environment for Northstar Health Plan. The environment is read-only and provides business APIs plus a SQL query endpoint. All tasks consume environment records and return structured JSON.1314## Quick start1516Every task shares the same environment access pattern. Use the base URL and credentials provided in the task prompt or `task_context.json` payload:1718```19Base URL: <TASK_ENV_BASE_URL>20SQL endpoint: POST {base_url}/sql/query21Auth header: Authorization: Bearer pa-review-token-01422```2324**Primary access method: SQL.** Use `POST /sql/query` for most data retrieval. The business REST endpoints are available for direct lookups by ID. See [schema.md](references/schema.md) for the full 19-table schema.2526Run SQL queries with the bundled script:2728```bash29python3 skill/scripts/query.py "<SQL>" --base-url <BASE_URL> --token pa-review-token-01430```3132Or directly with curl:3334```bash35curl -s -X POST <BASE_URL>/sql/query \36 -H "Authorization: Bearer pa-review-token-014" \37 -H "Content-Type: application/json" \38 -d '{"sql": "<SQL>"}'39```4041## Core workflow42431. **Discover the target**: Identify the case, claim, appeal, or queue ID from the prompt. Cross-reference against `task_context.json` when present.44452. **Query the environment**: Pull all relevant records. Start broad (the target record joined with its dependents), then narrow with WHERE clauses.46473. **Evaluate against criteria**: When a policy applies, match `policy_criteria` against `case_criteria` results and supporting `document_facts`. Check `is_current` on documents to separate active evidence from stale.48494. **Apply the source-precedence rule**: Every task has a `source_precedence` rule that controls how conflicting or overlapping records are resolved. See [precedence.md](references/precedence.md).50515. **Build the answer**: Fill the template fields from `answer_template.json`. Currency values rounded to 2 decimal places. Ratios to 4 decimal places. Null for absent modifiers, not empty string. ISO 8601 dates. Lists ordered as specified by the template.52536. **Construct the basis audit**: Every answer ends with a `basis_audit` containing the source-precedence rule, the ordered record precedence trail, the controlling record IDs, and exception/gap record IDs. See [precedence.md](references/precedence.md).5455## Task domains5657The environment supports five task domains, each with specific data models and evaluation patterns. Detailed workflows are in [workflows.md](references/workflows.md).5859| Domain | Typical Entry | Key Tables | Precedence Rule |60|---|---|---|---|61| UM nurse prior-auth | case_id | cases, request_lines, documents, document_facts, policy_criteria, case_criteria, members, plans, authorizations | current_clinical_records_over_stale_export |62| Pharmacy appeals | appeal_id | appeals, cases, drug_trials, assistance_screen, documents, policy_criteria | payer_appeal_before_manufacturer_assistance |63| Payment integrity | claim_id | claims, claim_lines, payment_benchmarks, plans | effective_benchmark_by_plan_modifier_and_date |64| Peer-to-peer | p2p_event_id | p2p_events, cases, request_lines, documents, policy_criteria | new_patient_specific_p2p_information |65| Margin queue | queue_id | service_margin | margin_threshold_then_charge_sensitivity |6667## Policies6869Five active policies cover the task domains. See [policies.md](references/policies.md) for policy-to-criterion mappings.7071| Policy ID | Name | Domains |72|---|---|---|73| POL-PT-LUMBAR-2026 | Lumbar Physical Therapy Medical Necessity | UM nurse prior-auth |74| POL-DRUG-EXC-2026 | Specialty Drug Coverage Exception Appeal | Pharmacy appeals |75| POL-PET-MPI-2026 | PET Myocardial Perfusion Imaging Medical Necessity | P2P |76| POL-CLAIM-RATE-2026 | Outpatient Imaging and Surgery Payment Benchmark | Payment integrity |77| POL-ST-PEDS-2026 | Pediatric Speech Therapy Prior Authorization | (not used in train tasks) |7879## Numeric conventions8081- **Currency** (USD): JSON number rounded to 2 decimal places.82- **Ratios** (e.g. revenue_to_cost_ratio): JSON number rounded to 4 decimal places.83- **Units** (e.g. approved_units, visits): integer.84- **Null modifier**: Use `null`, not empty string.85- **Dates**: ISO 8601 `YYYY-MM-DD`. Months as `YYYY-MM`.8687## Reference files8889- [schema.md](references/schema.md) — Full database schema with all 19 tables, columns, types, and keys.90- [precedence.md](references/precedence.md) — Source-precedence rules and the basis_audit construction pattern.91- [policies.md](references/policies.md) — Policy framework with criteria-key to criterion-ID mappings.92- [workflows.md](references/workflows.md) — Step-by-step workflows for each of the five task domains.