Licensing Review
This skill covers three regulatory licensing review workflows:
- Contractor eligibility batch review — cross-reference applications against bonds, insurance, license history, violations, inspections, and correspondence to produce APPROVE / HOLD / DENY decisions.
- Liquor license staff package — build a single-application review package with risk coverage analysis, verification gaps, monitoring plans, and escalation triggers.
- Alcohol renewal review queue — rank licensees for manual review using violation matching against a boundary date.
All workflows share a common environment pattern. Every prompt provides <TASK_ENV_BASE_URL> and the relevant API endpoints. Use the references/api_patterns.md reference for the standard fetch pattern, credential, and SQL access rules.
Universal Environment Setup
Before any domain-specific work, read the environment_access.md file staged alongside the task for the base URL, allowed endpoints, and the X-Task-Token credential value. Note the following:
- All GET endpoints return JSON arrays. Fetch them in parallel when they are independent.
POST /api/sqlaccepts{"query": "<SQL>"}with the headerX-Task-Token: <token-value>. Use it for cross-dataset joins or filtered lookups when the GET endpoints alone cannot disambiguate records.- The
<TASK_ENV_BASE_URL>placeholder in the prompt must be replaced with the actual base URL fromenvironment_access.md. - Templates provided under
input/payloads/answer_template.jsondefine the exact output structure, allowed enum values, ordering rules, and empty-value conventions. Never deviate from the template's schema.
Domain 1: Contractor Eligibility Batch Review
This domain covers batch-review tasks: reviewing a batch of contractor applications with IDs like C-XXXX-###.
Data Sources
Fetch all of these in parallel:
| Endpoint | What it provides |
|---|---|
GET /api/policies |
Current policy baselines, effective dates, new 2025 standards |
GET /api/contractor/applications |
Application records (endorsements, experience claims, status) |
GET /api/contractor/bonds |
Bond amounts, statuses, effective/cancellation dates |
GET /api/contractor/insurance |
Insurance coverage amounts, expiry dates, carrier status |
GET /api/contractor/license-history |
Prior licenses, suspensions, revocations |
GET /api/contractor/violations |
Violation records with severity and resolution status |
GET /api/contractor/correspondence |
Communication records, stale/unverified flags |
GET /api/contractor/inspections |
Inspection reports, safety flags, documentation gaps |
POST /api/sql |
Use when a record needs filtering by application ID that the GET endpoint doesn't directly support |
Decision Procedure
For each application in the batch, inspect these dimensions in order:
1. License History Check
Look at /api/contractor/license-history for the applicant. If any entry shows an active_suspension, assign active_suspension as a deficiency code and record risk_tier: "high". An active suspension alone is sufficient to recommend DENY.
2. Bond Check
Look at /api/contractor/bonds for the applicant. Check:
- Is there an active bond on file? If not, assign
no_active_bondorbond_cancelleddepending on whether a bond existed but was cancelled. - Is the bond amount sufficient compared to the policy-required minimum? If short, assign
bond_shortfall. - Is the bond expired or cancelled as of the review date? If so, assign
bond_cancelledorbond_shortfall.
3. Insurance Check
Compare each insurance record against the review date given in the prompt:
- Expired as of review date →
insurance_expired - Not yet effective / pending →
insurance_pending - Coverage amount below policy minimum →
insurance_shortfall - No record found or coverage gap →
insurance_not_current
4. Endorsement Check
From the application record and any related data:
- Required endorsement not obtained →
endorsement_missing - Application in progress →
endorsement_pending - Status cannot be verified →
endorsement_not_verified
5. Experience Check
From the application record:
- Insufficient documented years or project count →
experience_shortfall
6. Violation / Complaint Check
- Unresolved serious violations →
open_serious_violation - Unresolved minor violations →
open_minor_violation - Unresolved serious complaints →
unresolved_serious_complaint
7. Inspection Check
- Missing required inspection documentation →
inspection_doc_gap - Failed safety inspection requiring recheck →
inspection_safety_recheck
Determination Rule
| Determination | Condition |
|---|---|
DENY |
active_suspension is present, OR open_serious_violation/unresolved_serious_complaint combined with multiple additional deficiencies from other categories |
APPROVE |
Zero deficiency codes found |
HOLD |
Deficiencies exist but none of the DENY triggers apply |
Risk Tier Assignment
| Tier | Condition |
|---|---|
low |
Zero deficiencies (APPROVE) |
high |
active_suspension, open_serious_violation, unresolved_serious_complaint, or deficiencies crossing three or more of the check categories above |
medium |
Other deficiency patterns |
Policy Impact
policy_impacted: true when the current policy baseline (GET /api/policies) introduces a 2025-or-later standard that creates a deficiency or material review flag that would not have applied under prior rules. Compare each deficiency against the policy effective dates and scope descriptions.
Required Actions
Map each deficiency code to required actions using the reference table in references/contractor_code_mappings.md. Sort actions alphabetically.
Summary Construction
After deciding all applications:
- Count APPROVE, HOLD, DENY.
- Collect all
highrisk tier application IDs, sorted ascending. - Collect all
policy_impacted: trueapplication IDs, sorted ascending. - Identify correspondence IDs with stale or unverified status from
/api/contractor/correspondence, sorted ascending.
Output
Produce only the JSON object matching the task's answer_template.json. Order application_decisions by application_id ascending. Use empty arrays when no codes apply.
Domain 2: Liquor License Staff Package
This domain covers single-application liquor license review tasks. The prompt names one target application ID and one target location ID.
Data Sources
Fetch all of these in parallel:
| Endpoint | What it provides |
|---|---|
GET /api/policies |
Policy rules, premises definitions, control requirements |
GET /api/liquor/applications |
Application details, license class, requested privileges |
GET /api/liquor/settlements |
Prior settlement agreements, board orders, consent decrees |
GET /api/liquor/privileges |
Active privilege grants and restrictions |
GET /api/liquor/incidents |
Police incidents, complaints, infractions at the location |
GET /api/liquor/site-evidence |
Floor plans, photos, CCTV specs, signage, food service evidence |
POST /api/sql |
Use for cross-referencing records when direct GET filtering is insufficient |
Decision Procedure
1. Same-Premises Basis
Check whether the application falls under a same-premises rule (previous license at this location, prior board findings about the site). Look at /api/liquor/settlements and /api/liquor/privileges for references to the target location. Set same_premises_basis_applies: true when prior license or settlement records reference the same physical premises.
2. Covered Risk Assessment
Review all incidents, settlements, and site evidence to identify risks that have adequate existing controls or documentation coverage. Examples of covered risks:
AFTER_HOURS— after-hours service risk with monitoring controlsASSAULT— assault/harassment incidents with security controlsMINOR_SALE/SALE_TO_MINOR— age-verification controlsNOISE/PATIO_BOUNDARY— noise or patio boundary with monitoringSAME_PREMISES— same-premises basis provides control coverageFOOD_SERVICE_GAP— food service requirements metTAX_HOLD— tax clearance obtained
The allowed codes are listed in the task's answer template. Pick only codes that are actually supported by evidence from the data.
3. Verification Gap Assessment
Identify what is missing, conflicting, or unverified. Common gap patterns:
- Signage documentation missing or conflicting →
CONTROL_SIGNAGE_CURRENT_MISSING,CONTROL_SIGNAGE_CONFLICTING - Floor plan stale or conflicting →
FLOOR_PLAN_CONFLICTING,FLOOR_PLAN_STALE - Open incident requiring follow-up →
OPEN_INCIDENT_FOLLOW_UP - Police memo with conflicting information →
POLICE_MEMO_CONFLICTING - Missing neighbor notice →
NEIGHBOR_NOTICE_MISSING - Missing site photos →
SITE_PHOTO_MISSING - Tax clearance missing →
TAX_CLEARANCE_MISSING - Camera evidence missing →
camera_evidence_missing - Food service evidence missing →
food_service_evidence_missing - Late-night monitoring needed →
late_night_monitoring_needed
Use only codes from the template's allowed values.
4. Standard vs Location-Specific Obligations
- Standard obligations: Required for all licensees of this class regardless of location. Usually include
ID_CHECK,HOURS,FOOD_SERVICE. Also includeCCTV,NOISE,SECURITY,PATIO,DELIVERYwhen the license class or jurisdiction mandates them universally. - Location-specific controls: Controls tied to this particular premises due to its history, layout, or settlement terms. Examples:
CCTVrequired by a settlement at this address,HOURSrestricted by a prior board order,SECURITYmandated by an incident pattern,NOISE/PATIOdue to outdoor service area.
5. Recommended Posture
| Posture | Condition |
|---|---|
deny |
Unresolvable risks, board order conflicts, or evidence of unfitness |
request_follow_up |
Verifiable gaps exist that the applicant can resolve with additional submissions |
issue_restricted |
Risks are covered by existing controls and only standard obligations apply |
Default to request_follow_up when gaps are present but resolvable.
6. 90-Day Monitoring Plan
Build a list of monitoring checks with timing slots. Each check has a check_code and a timing:
first_30_days: Immediate post-issuance verification (signage, ID check observation, camera export test, food service check, police memo follow-up, security walkthrough, tax clearance)days_31_60: Mid-period checks (after-hours visits, late-night closing visits)days_61_90: Late-period checks (noise/patio boundary, log reviews)
Select checks that correspond to the identified verification gaps and location-specific controls. Sort by operational sequence: immediate verifications first, then monitoring visits, then boundary checks.
7. Escalation Triggers
Identify conditions that should cause field staff to escalate. Map each major risk and verification gap to a trigger code from the template. Examples:
- After-hours risk →
AFTER_HOURS_VIOLATION - Signage unverified →
CONTROL_SIGNAGE_NOT_VERIFIED - Major incident →
MAJOR_INCIDENT_REPORTED - Minor sale unresolved →
REFERRED_MINOR_SALE_UNRESOLVED - CCTV control failure →
SECURITY_CCTV_CONTROL_FAILURE - Tax hold reopened →
TAX_HOLD_REOPENED - Camera coverage gap →
missing_camera_coverageorfootage_not_produced - Food service gap →
food_service_not_available - Noise/patio breach →
noise_or_patio_breach - Tax hold uncleared →
open_tax_hold_uncleared
The full code-to-evidence reference table is in references/liquor_code_mappings.md.
Output
Produce only the JSON object matching the answer template schema. Include exactly the required top-level keys. Sort code arrays ascending (or as specified by the template's ordering rules). Remove duplicates.
Domain 3: Alcohol Renewal Review Queue
This domain covers building a ranked manual-review queue for alcohol license renewals.
Data Sources
| Endpoint | What it provides |
|---|---|
GET /api/alcohol/licensees |
License records with facility names, addresses, license numbers |
GET /api/alcohol/violations |
Violation records with dates, descriptions, license references |
GET /api/renewal/rules |
Renewal rules, boundary dates, ranking criteria |
POST /api/sql |
Use for joining licensee and violation data or filtering by date range |
Decision Procedure
1. Boundary Date Filtering
The prompt specifies a boundary date (e.g., 2025-04-10). Violations dated after this boundary are excluded from the ranking calculation. Collect these excluded IDs for the summary.
2. Violation Matching
For each target license, match violations from /api/alcohol/violations:
- Exact match: violation record references the license number directly.
- Close address match: violation record references a different license number but the facility address is substantially similar (same street, nearby unit).
- Uncertain match: the connection is plausible but cannot be confirmed from available data alone.
Use POST /api/sql when the violation records use non-obvious identifiers that require a join with licensee data.
3. Queue Ranking
Rank licenses by a composite that prioritizes:
- Violation recency: violations within the final ~30 days before the boundary date carry the most weight.
- Violation count: more pre-boundary violations rank higher.
- Match confidence: exact matches rank above close-address or uncertain matches when other factors are equal.
The ranking should place the most inspection-urgent licenses at the top. Licenses with recent violations just before the boundary are the highest priority.
4. Risk Tier and Next-Step Labels
| Risk Tier | Typical Assignment |
|---|---|
high |
Any matched pre-boundary violations, or close/uncertain matches suggesting additional risk |
medium |
Lower violation counts with older dates |
low |
No violations or only very old, minor ones |
| Next-Step Label | Typical Assignment |
|---|---|
board_review |
Highest-priority entries (top of queue), multiple recent violations |
manual_fine_check |
Mid-priority entries with moderate violation counts |
manual_ALERT_check |
Lower-priority entries |
additional_record_check |
Entries needing further data verification |
5. Summary Construction
queue_size: number of entries in the queue (usually the target queue size from the prompt)boundary_date: the boundary date from the promptpost_boundary_violation_ids_excluded: all violation IDs dated after the boundary, sorted ascendingclose_or_uncertain_match_license_numbers: license numbers with non-exact matches, sorted ascendingboard_review_license_numbers: license numbers assignedboard_review, sorted ascending
Output
Produce the queue list with ascending ranks (1 through N, no gaps), ordered by rank. Each entry must include all fields from the answer template. Sort matched_violation_ids by violation date ascending, then by ID ascending.
Cross-Domain Principles
- Ordering: Sort string lists ascending, integer ranks ascending, and dates chronologically unless the template says otherwise.
- Empty values: Use
[]for empty arrays, nevernullor omitted keys. - No prose: Return only the JSON object. Do not include narrative memos, citations, markdown wrapping, or additional commentary.
- Template compliance: The answer template is authoritative for allowed enum values and key names. Even if the data suggests a code not in the template, do not invent new codes. If a condition exists but no template code covers it, note it in the most applicable existing code or omit it.
- SQL usage: Use
POST /api/sqlwhen GET endpoints return broad datasets that need filtering, or when records need cross-joining. Always pass the token header.