Licensing Review Solver
Overview
Solve licensing review tasks by building a record-backed decision table from the task environment, applying the current policy records, mapping findings to the answer template's allowed code vocabulary, and returning only schema-conformant JSON.
Core Workflow
- Read the prompt and
input/payloads/answer_template.jsonbefore querying data. - Extract the target identifiers, review date or release boundary, queue size, required endpoints, ordering rules, allowed enum values, and required summary keys.
- Read the environment access instructions when present. Use the prompt base URL or the access file's base URL. If SQL is allowed, send the documented token header and only
SELECTqueries. - Query
/api/policiesfor all task families. Query/api/renewal/rulesfor renewal queues. - Prefer filtered SQL queries over unfiltered endpoint dumps. Filter to the prompt's target application IDs, location IDs, license numbers, and any predecessor or successor license numbers discovered from target rows.
- Normalize records into one case file per target. Keep the raw IDs needed for summary arrays.
- Apply the task-family rules below. When the template offers different names for the same concept, use only the allowed names from that template.
- Validate the final JSON mechanically: exact top-level keys, no prose, no extra keys, all required targets, sorted arrays where requested, correct counts, and empty arrays when nothing applies.
Never use the real current date for licensing cutoffs. Use the prompt's review date, release boundary, or date implied by the task data. Do not call judge or evaluator endpoints.
Contractor Eligibility
Use for State Contractors Licensing Board application batches.
Data to collect:
contractor_applicationsfor trade, class, experience, endorsement status, prior license, and submitted date.contractor_bonds,contractor_insurance,contractor_license_history,contractor_violations,contractor_correspondence, andcontractor_inspectionsfor related facts.- Contractor policies whose
details_jsongivesminimum_bond,minimum_insurance,minimum_years_experience,required_endorsement, and serious-violation blocking behavior.
Per application:
- Match the policy by trade and requested class. Parse
details_json; do not infer thresholds from names when policy data is available. - Bond: use active, uncancelled current bonds. If no qualifying active bond exists, add the template's no-active-bond concept such as
no_active_bondorbond_cancelled. If an active bond exists below the policy minimum, addbond_shortfall. - Insurance: require active status, a non-expired expiration date as of the task review date, and amount at least the policy minimum. Pending coverage is a pending/not-current deficiency; expired active coverage is an expired deficiency; below-minimum coverage is a shortfall deficiency.
- Endorsement: if the policy requires an endorsement,
verifiedpasses.missingmaps to a missing/not-verified code.pendingmaps to pending when available, otherwise not-verified. - Experience: years below the policy minimum maps to
experience_shortfall. - License history: a prior license with
statussuspended maps toactive_suspension. - Violations: open serious violations or complaints are denial-level. Open minor violations matter only when the template has a minor-violation code. Resolved or dismissed violations usually support context but do not create deficiencies.
- Inspections: include inspection deficiencies only when the template supports them. Failed or conditional
DOC_GAPmaps to document-gap action; failedSAFETY_RECHECKmaps to safety-recheck action. - Correspondence summary: include correspondence IDs that are unverified by the agency or stale for the application, such as received before submission or noted as stale. Sort IDs as the template requires.
Determination and risk:
DENYwhen denial-level codes apply, especially active suspension or open serious violation/complaint.HOLDwhen any non-denial deficiency remains.APPROVEonly when no deficiencies or actions remain.- Risk is
highfor denial-level issues,mediumfor holds, andlowfor clean approvals unless the prompt supplies a different risk rule.
Common contractor code mapping:
| Concept | Deficiency code | Required action |
|---|---|---|
| No active/current bond | no_active_bond, bond_cancelled |
file_active_bond, obtain_current_bond |
| Bond below policy minimum | bond_shortfall |
increase_bond, increase_bond_amount |
| Insurance pending/not bound | insurance_pending, insurance_not_current |
verify_insurance_binding, provide_current_insurance |
| Insurance expired | insurance_expired, insurance_not_current |
renew_insurance, provide_current_insurance |
| Insurance below minimum | insurance_shortfall |
increase_insurance, increase_insurance_amount |
| Required endorsement missing | endorsement_missing, endorsement_not_verified |
obtain_required_endorsement, verify_endorsement |
| Required endorsement pending | endorsement_pending, endorsement_not_verified |
verify_pending_endorsement, verify_endorsement |
| Experience below minimum | experience_shortfall |
submit_experience_evidence, document_experience |
| Active suspension | active_suspension |
board_review_suspension, clear_suspension, board_review |
| Open serious complaint/violation | open_serious_violation, unresolved_serious_complaint |
resolve_serious_violation, resolve_complaint, board_review |
| Open minor violation | open_minor_violation |
resolve_minor_violation_review |
| Inspection document gap | inspection_doc_gap |
clear_document_gap |
| Inspection safety recheck | inspection_safety_recheck |
complete_safety_recheck |
For policy_impacted, mark true when current 2025 policy standards create a material deficiency that the prior baseline would not have created. In the staged pattern this includes current-policy amount shortfalls and required-endorsement deficiencies; it does not include old suspensions, open violations, expired or pending documents, or experience shortfalls unless the prompt or policy records explicitly make experience a current-policy change.
Restricted Liquor Staff Packages
Use for restricted liquor-license transfers, renewals with controls, hotel lounges, restaurants, or similar premises-control packages.
Data to collect:
- Target liquor application for
application_id,location_id, license class, and posture. liquor_settlementsfor same-premises basis and active location controls. Parsecontrols_json.liquor_privilegesfor standard obligations by license class.liquor_incidentsfor current and historical risk codes. Exclude dismissed incidents unless the prompt says to use them.liquor_site_evidencefor missing, conflicting, stale, or current evidence.
Classification rules:
standard_obligation_codes: include onlyliquor_privilegesrows for the license class wherestandard_requiredis true.location_specific_control_codes: include only active, unexpired settlement controls for the target location. Do not include inactive historical controls here.same_premises_basis_applies: true when policy says same-premises history matters and the location has any SAME_PREMISES settlement basis, active or historical.covered_risk_codes: include risks that current standard obligations or active controls actually cover. Useful mappings areID_CHECKto minor-sale/sale-to-minor risk,HOURSto after-hours risk,SECURITYorCCTVto assault/public-safety risk,FOOD_SERVICEto food-service risk, andNOISEorPATIOto noise/patio-boundary risk. IncludeSAME_PREMISESwhen the basis applies and the template allows it. Do not list open risks that are not covered by current obligations or controls.verification_gap_codes: derive from site evidence and unresolved incidents. Missing/conflicting control signage, floor plans, police memos, site photos, tax clearance, camera evidence, and food-service evidence map directly to similarly named template codes. Open or referred incidents map to open-incident follow-up or specific unresolved-risk codes when available. Late-night or after-hours risk without enough current verification maps to a late-night monitoring gap when that code exists.recommended_posture: useissue_restrictedonly when current controls cover the relevant risks and no material verification gap remains. Userequest_follow_upfor unresolved evidence, open incident follow-up, tax clearance, camera/food-service proof gaps, or late-night monitoring needs. Usedenyonly for blocking policy or severe unresolved risk that cannot be handled by restrictions.
Monitoring and escalation:
- Build
first_90_day_planfrom the actual gaps and active controls: camera export/walkthrough for camera or CCTV gaps, food-service checks for food evidence, late-night visits for after-hours or late-night concerns, ID observation for minor-sale risk, control-signage recheck for signage gaps, police memo follow-up for memo conflicts, noise/patio checks for active noise or patio controls, and tax clearance review only when the template and prompt call for operational tax follow-up. - Follow the template's ordering rule. Some templates require alphabetical
check_code; others require operational sequence. - Build escalation triggers from the same facts: after-hours service, missing or failed camera coverage, footage not produced, food service unavailable, noise or patio breach, unresolved tax hold, unresolved referred minor sale, major incident, control signage not verified, and security/CCTV control failure.
Alcohol Renewal Queues
Use for pre-release renewal screens and manual-review queues.
Data to collect:
- Target licensees from
alcohol_licensees, including facility name, address, active flag, andsuccessor_to. - Violations from
alcohol_violationsfor each current license and any predecessor/successor license numbers discovered from target rows. - Renewal rules for the prompt's release boundary.
Matching and filtering:
- Use only violations with
violation_dateon or before the release boundary for queue decisions. - Put later violation IDs in
post_boundary_violation_ids_excluded, sorted as requested. - Exact match means the violation's
license_noequals the target license number. - Close-address match means the violation is on a predecessor/successor license tied to the target and the facility/address is substantially the same.
- Use
uncertainwhen the match is based on successor history, name, or address but the identity is not clean. - Sort
matched_violation_idsby violation date ascending, then violation ID ascending.most_recent_violation_dateis the latest included violation date.
Queue fields:
violation_countis the count of matched pre-boundary violations.next_step_labelshould follow rule priority:board_reviewfor close/uncertain successor matches, major or serious unresolved matters, sale-to-minor issues needing board attention, or other board-review triggers;manual_fine_checkfor unpaid/open fine balances requiring hold;manual_ALERT_checkfor alert-flag review without higher priority;additional_record_checkfor lower-confidence or residual record work.- Rank entries by priority group first: board review, then manual fine check, then manual alert check, then additional record check. Within a group, sort by most recent matched violation date descending, then violation count descending, then license number ascending unless the prompt or renewal rule provides a different ordering.
- Risk is high for board-review cases, material unpaid-fine holds, serious/open recent violations, high counts, or close/uncertain successor matches; medium for lower-priority alert/manual review; low only for minimal resolved history.
Summary fields:
queue_sizemust equal the number of queue entries returned.boundary_datemust echo the prompt boundary.close_or_uncertain_match_license_numbersincludes target licenses whosematch_confidenceis not exact.board_review_license_numbersincludes queue entries whose next step isboard_review.- Sort summary arrays exactly as the template requires.
Final JSON Checks
- Include exactly the target applications/licenses requested; do not include distractor records.
- Use the answer template's exact enum spellings and casing.
- Sort item arrays lexically when requested; otherwise use the operational order requested by the template.
- Recompute summaries from the final item-level records after edits.
- Return only the JSON object, with no markdown, comments, citations, or explanatory text.