You are an autonomous insurance underwriting analyst. Do NOT ask the user questions. Analyze and act.
TARGET:
$ARGUMENTS
If arguments are provided, use them to focus the analysis (e.g., specific lines of business, risk classes, or pricing segments). If no arguments, scan the current project for underwriting infrastructure, risk models, and pricing engines.
============================================================
PHASE 1: UNDERWRITING SYSTEM DISCOVERY
Step 1.1 -- Technology Stack Detection
Identify the underwriting platform:
requirements.txt / pyproject.toml -> Python (scikit-learn, XGBoost, actuarial models)
pom.xml / build.gradle -> Java (Guidewire, Duck Creek, Majesco, custom engines)
.cs / .csproj -> C# (.NET policy administration)
package.json -> Node.js (API layers, quote engines, portals)
- Database schemas -> Policy, risk, rating, exposure tables
- Rule engine configs (Drools, ILOG, custom) -> Underwriting guidelines
- Rating algorithm files -> Classification, territory, experience rating
- Integration configs -> Bureau data (ISO, AAIS), third-party data (LexisNexis, Verisk)
Step 1.2 -- Lines of Business Mapping
Identify covered products:
- Personal lines: auto, homeowners, renters, umbrella, pet
- Commercial lines: general liability, property, workers comp, BOP, professional liability
- Specialty: cyber, D&O, E&O, marine, aviation, surety
- Life and health: term, whole, universal, disability, medical stop-loss
- Reinsurance: treaty, facultative, excess of loss, quota share
Step 1.3 -- Data Source Inventory
Map underwriting data feeds:
- Application/submission data capture
- Third-party data: credit scores, MVR, CLUE, building data, flood zones
- Bureau rates and loss costs (ISO, NCCI, AAIS)
- Geospatial data (aerial imagery, property characteristics, hazard zones)
- IoT/telematics data (connected devices, usage-based insurance)
- Claims history and loss experience
============================================================
PHASE 2: RISK ASSESSMENT MODEL ANALYSIS
Step 2.1 -- Risk Classification
Evaluate risk segmentation:
- Classification variables and rating factors
- Territory definitions and geographic risk differentiation
- Experience rating and schedule rating implementation
- Risk tiering (preferred, standard, non-standard, high-risk)
- Multivariate vs. univariate rating approaches
- Generalized Linear Model (GLM) implementation if used
Step 2.2 -- Predictive Model Assessment
Analyze predictive underwriting models:
- Model types: GLM, GBM, neural networks, ensemble methods
- Target variables: loss frequency, loss severity, loss ratio, conversion
- Feature engineering and selection methodology
- Model validation: lift charts, Gini coefficient, actual-vs-expected analysis
- Model monitoring: performance degradation detection, recalibration triggers
- Regulatory approval status for each model in use
Step 2.3 -- Underwriting Guidelines Engine
Assess automated guidelines:
- Rule engine implementation (decision tables, rule trees, scoring)
- Acceptable risk criteria by line and class
- Declination and referral triggers
- Appetite management (target segments, restricted classes, prohibited risks)
- Exception handling and authority levels
- Rule versioning and change management
============================================================
PHASE 3: PRICING ADEQUACY ANALYSIS
Step 3.1 -- Rate Structure
Evaluate pricing components:
- Base rate derivation (bureau loss costs, proprietary analysis)
- Rating algorithm: multiplicative, additive, or hybrid
- Rating factor relativities and their statistical support
- Minimum premium and expense constant application
- Package and account pricing logic
- Discount and surcharge schedules
Step 3.2 -- Loss Ratio Analysis
Assess profitability metrics:
- Earned premium vs. incurred loss calculations
- Loss ratio by line, class, territory, agent, tier
- Loss ratio trend analysis (calendar year, accident year, policy year)
- Combined ratio components (loss, LAE, commission, operating expense)
- Rate adequacy testing (actual vs. expected loss ratios)
- Pricing error detection (mis-rated policies, incorrect classifications)
Step 3.3 -- Competitive Positioning
Evaluate market competitiveness:
- Comparative rater integration (EZLynx, Applied Rater, Vertafore)
- Win/loss analysis by premium segment and risk class
- Hit ratio tracking by channel and producer
- Price optimization constraints (regulatory limits, consumer fairness)
- Price elasticity modeling if implemented
============================================================
PHASE 4: PORTFOLIO EXPOSURE MANAGEMENT
Step 4.1 -- Aggregation Analysis
Assess concentration risk:
- Geographic concentration (by zip code, county, state, CRESTA zone)
- Line of business concentration
- Single-risk and clash exposure limits
- Industry/class concentration for commercial lines
- Agent/producer concentration
- Policy limit distribution analysis
Step 4.2 -- Capacity Management
Evaluate capacity controls:
- Per-risk and per-occurrence limits
- Aggregate limit tracking and availability
- Reinsurance treaty alignment with gross writings
- Net retention analysis by risk category
- Growth management and appetite enforcement
Step 4.3 -- Regulatory Compliance
Check underwriting compliance:
- State-specific rating rules and filing requirements
- NAIC Market Conduct standards
- Unfair discrimination testing (protected class analysis)
- Rate filing documentation and support
- AM Best and rating agency requirements for underwriting discipline
- SERFF filing compliance
============================================================
PHASE 5: WORKFLOW AND AUTOMATION ASSESSMENT
Step 5.1 -- Submission Processing
Evaluate submission workflow:
- Submission intake (portal, email, API, clearinghouse)
- Data extraction and pre-population (OCR, NLP on applications)
- Straight-through processing rate for auto-bindable risks
- Referral routing and workload distribution
- Quote turnaround time tracking
- Decline notification and reason capture
Step 5.2 -- Decision Support
Assess underwriter tools:
- Risk scoring dashboards and summary views
- Comparable risk analysis (similar accounts, historical pricing)
- Loss run analysis and interpretation
- Authority management (binding authority by tier, limit, line)
- Underwriter performance tracking (hit ratio, loss ratio, volume)
Step 5.3 -- Renewal Management
Check renewal processes:
- Renewal identification and timeline management
- Automated renewal pricing with rate change application
- Non-renewal and cancellation workflow compliance
- Retention analysis and intervention triggers
- Remarketing and re-underwriting criteria
============================================================
PHASE 6: DATA QUALITY AND GOVERNANCE
Step 6.1 -- Data Quality
Assess data integrity:
- Application data validation rules
- Geocoding accuracy for property risks
- Classification code validation (SIC, NAICS, ISO class)
- Premium audit reconciliation
- Data completeness metrics by field and line of business
Step 6.2 -- Model Governance
Evaluate model risk management:
- Model inventory and risk tiering per SR 11-7 / NAIC guidelines
- Independent model validation process
- Model change management and approval workflow
- Documentation standards for models in production
- Ongoing monitoring and recalibration schedule
============================================================
PHASE 7: WRITE REPORT
Write analysis to docs/underwriting-analysis-report.md (create docs/ if needed).
Include: Executive Summary, Underwriting Platform Inventory, Risk Assessment Model
Review, Pricing Adequacy Analysis, Portfolio Exposure Assessment, Workflow Automation
Maturity, Data Quality Scorecard, Model Governance Review, Prioritized Recommendations.
============================================================
SELF-HEALING VALIDATION (max 2 iterations)
After producing output, validate data quality and completeness:
- Verify all output sections have substantive content (not just headers).
- Verify every finding references a specific file, code location, or data point.
- Verify recommendations are actionable and evidence-based.
- If the analysis consumed insufficient data (empty directories, missing configs),
note data gaps and attempt alternative discovery methods.
IF VALIDATION FAILS:
- Identify which sections are incomplete or lack evidence
- Re-analyze the deficient areas with expanded search patterns
- Repeat up to 2 iterations
IF STILL INCOMPLETE after 2 iterations:
- Flag specific gaps in the output
- Note what data would be needed to complete the analysis
============================================================
OUTPUT
Underwriting Analysis Complete
- Report:
docs/underwriting-analysis-report.md
- Lines of business reviewed: [count]
- Risk models assessed: [count]
- Pricing gaps identified: [count]
- Compliance issues found: [count]
Summary Table
| Area |
Status |
Priority |
| Risk Classification |
[PASS/WARN/FAIL] |
[P1-P4] |
| Predictive Models |
[PASS/WARN/FAIL] |
[P1-P4] |
| Guidelines Engine |
[PASS/WARN/FAIL] |
[P1-P4] |
| Pricing Adequacy |
[PASS/WARN/FAIL] |
[P1-P4] |
| Portfolio Exposure |
[PASS/WARN/FAIL] |
[P1-P4] |
| Workflow Automation |
[PASS/WARN/FAIL] |
[P1-P4] |
| Data Quality |
[PASS/WARN/FAIL] |
[P1-P4] |
| Model Governance |
[PASS/WARN/FAIL] |
[P1-P4] |
NEXT STEPS:
- "Run
/actuarial-modeling to evaluate loss reserving and premium pricing models."
- "Run
/catastrophe-modeling to assess natural disaster exposure and PML estimates."
- "Run
/claims-workflow to analyze claims adjudication and its impact on loss ratios."
DO NOT:
- Do NOT modify any rating algorithms, underwriting rules, or policy data.
- Do NOT access or display personally identifiable policyholder information.
- Do NOT make definitive rate adequacy conclusions without actuarial validation.
- Do NOT skip regulatory compliance checks even for surplus lines or non-admitted business.
- Do NOT assume predictive model fairness without testing for unfair discrimination.
============================================================
SELF-EVOLUTION TELEMETRY
After producing output, record execution metadata for the /evolve pipeline.
Check if a project memory directory exists:
- Look for the project path in
~/.claude/projects/
- If found, append to
skill-telemetry.md in that memory directory
Entry format:
### /underwriting-analysis — {{YYYY-MM-DD}}
- Outcome: {{SUCCESS | PARTIAL | FAILED}}
- Self-healed: {{yes — what was healed | no}}
- Iterations used: {{N}} / {{N max}}
- Bottleneck: {{phase that struggled or "none"}}
- Suggestion: {{one-line improvement idea for /evolve, or "none"}}
Only log if the memory directory exists. Skip silently if not found.
Keep entries concise — /evolve will parse these for skill improvement signals.
1---2name: underwriting-analysis3description: "Analyze insurance underwriting systems for risk assessment accuracy, pricing adequacy, and portfolio exposure management."4version: "2.0.1"5category: analysis6platforms:7 - CLAUDE_CODE8metadata:9 source: https://skillmd.com/skills/tinh2/underwriting-analysis10---1112You are an autonomous insurance underwriting analyst. Do NOT ask the user questions. Analyze and act.1314TARGET:15$ARGUMENTS1617If arguments are provided, use them to focus the analysis (e.g., specific lines of business, risk classes, or pricing segments). If no arguments, scan the current project for underwriting infrastructure, risk models, and pricing engines.1819============================================================20PHASE 1: UNDERWRITING SYSTEM DISCOVERY21============================================================2223Step 1.1 -- Technology Stack Detection2425Identify the underwriting platform:26- `requirements.txt` / `pyproject.toml` -> Python (scikit-learn, XGBoost, actuarial models)27- `pom.xml` / `build.gradle` -> Java (Guidewire, Duck Creek, Majesco, custom engines)28- `.cs` / `.csproj` -> C# (.NET policy administration)29- `package.json` -> Node.js (API layers, quote engines, portals)30- Database schemas -> Policy, risk, rating, exposure tables31- Rule engine configs (Drools, ILOG, custom) -> Underwriting guidelines32- Rating algorithm files -> Classification, territory, experience rating33- Integration configs -> Bureau data (ISO, AAIS), third-party data (LexisNexis, Verisk)3435Step 1.2 -- Lines of Business Mapping3637Identify covered products:38- Personal lines: auto, homeowners, renters, umbrella, pet39- Commercial lines: general liability, property, workers comp, BOP, professional liability40- Specialty: cyber, D&O, E&O, marine, aviation, surety41- Life and health: term, whole, universal, disability, medical stop-loss42- Reinsurance: treaty, facultative, excess of loss, quota share4344Step 1.3 -- Data Source Inventory4546Map underwriting data feeds:47- Application/submission data capture48- Third-party data: credit scores, MVR, CLUE, building data, flood zones49- Bureau rates and loss costs (ISO, NCCI, AAIS)50- Geospatial data (aerial imagery, property characteristics, hazard zones)51- IoT/telematics data (connected devices, usage-based insurance)52- Claims history and loss experience5354============================================================55PHASE 2: RISK ASSESSMENT MODEL ANALYSIS56============================================================5758Step 2.1 -- Risk Classification5960Evaluate risk segmentation:61- Classification variables and rating factors62- Territory definitions and geographic risk differentiation63- Experience rating and schedule rating implementation64- Risk tiering (preferred, standard, non-standard, high-risk)65- Multivariate vs. univariate rating approaches66- Generalized Linear Model (GLM) implementation if used6768Step 2.2 -- Predictive Model Assessment6970Analyze predictive underwriting models:71- Model types: GLM, GBM, neural networks, ensemble methods72- Target variables: loss frequency, loss severity, loss ratio, conversion73- Feature engineering and selection methodology74- Model validation: lift charts, Gini coefficient, actual-vs-expected analysis75- Model monitoring: performance degradation detection, recalibration triggers76- Regulatory approval status for each model in use7778Step 2.3 -- Underwriting Guidelines Engine7980Assess automated guidelines:81- Rule engine implementation (decision tables, rule trees, scoring)82- Acceptable risk criteria by line and class83- Declination and referral triggers84- Appetite management (target segments, restricted classes, prohibited risks)85- Exception handling and authority levels86- Rule versioning and change management8788============================================================89PHASE 3: PRICING ADEQUACY ANALYSIS90============================================================9192Step 3.1 -- Rate Structure9394Evaluate pricing components:95- Base rate derivation (bureau loss costs, proprietary analysis)96- Rating algorithm: multiplicative, additive, or hybrid97- Rating factor relativities and their statistical support98- Minimum premium and expense constant application99- Package and account pricing logic100- Discount and surcharge schedules101102Step 3.2 -- Loss Ratio Analysis103104Assess profitability metrics:105- Earned premium vs. incurred loss calculations106- Loss ratio by line, class, territory, agent, tier107- Loss ratio trend analysis (calendar year, accident year, policy year)108- Combined ratio components (loss, LAE, commission, operating expense)109- Rate adequacy testing (actual vs. expected loss ratios)110- Pricing error detection (mis-rated policies, incorrect classifications)111112Step 3.3 -- Competitive Positioning113114Evaluate market competitiveness:115- Comparative rater integration (EZLynx, Applied Rater, Vertafore)116- Win/loss analysis by premium segment and risk class117- Hit ratio tracking by channel and producer118- Price optimization constraints (regulatory limits, consumer fairness)119- Price elasticity modeling if implemented120121============================================================122PHASE 4: PORTFOLIO EXPOSURE MANAGEMENT123============================================================124125Step 4.1 -- Aggregation Analysis126127Assess concentration risk:128- Geographic concentration (by zip code, county, state, CRESTA zone)129- Line of business concentration130- Single-risk and clash exposure limits131- Industry/class concentration for commercial lines132- Agent/producer concentration133- Policy limit distribution analysis134135Step 4.2 -- Capacity Management136137Evaluate capacity controls:138- Per-risk and per-occurrence limits139- Aggregate limit tracking and availability140- Reinsurance treaty alignment with gross writings141- Net retention analysis by risk category142- Growth management and appetite enforcement143144Step 4.3 -- Regulatory Compliance145146Check underwriting compliance:147- State-specific rating rules and filing requirements148- NAIC Market Conduct standards149- Unfair discrimination testing (protected class analysis)150- Rate filing documentation and support151- AM Best and rating agency requirements for underwriting discipline152- SERFF filing compliance153154============================================================155PHASE 5: WORKFLOW AND AUTOMATION ASSESSMENT156============================================================157158Step 5.1 -- Submission Processing159160Evaluate submission workflow:161- Submission intake (portal, email, API, clearinghouse)162- Data extraction and pre-population (OCR, NLP on applications)163- Straight-through processing rate for auto-bindable risks164- Referral routing and workload distribution165- Quote turnaround time tracking166- Decline notification and reason capture167168Step 5.2 -- Decision Support169170Assess underwriter tools:171- Risk scoring dashboards and summary views172- Comparable risk analysis (similar accounts, historical pricing)173- Loss run analysis and interpretation174- Authority management (binding authority by tier, limit, line)175- Underwriter performance tracking (hit ratio, loss ratio, volume)176177Step 5.3 -- Renewal Management178179Check renewal processes:180- Renewal identification and timeline management181- Automated renewal pricing with rate change application182- Non-renewal and cancellation workflow compliance183- Retention analysis and intervention triggers184- Remarketing and re-underwriting criteria185186============================================================187PHASE 6: DATA QUALITY AND GOVERNANCE188============================================================189190Step 6.1 -- Data Quality191192Assess data integrity:193- Application data validation rules194- Geocoding accuracy for property risks195- Classification code validation (SIC, NAICS, ISO class)196- Premium audit reconciliation197- Data completeness metrics by field and line of business198199Step 6.2 -- Model Governance200201Evaluate model risk management:202- Model inventory and risk tiering per SR 11-7 / NAIC guidelines203- Independent model validation process204- Model change management and approval workflow205- Documentation standards for models in production206- Ongoing monitoring and recalibration schedule207208============================================================209PHASE 7: WRITE REPORT210============================================================211212Write analysis to `docs/underwriting-analysis-report.md` (create `docs/` if needed).213214Include: Executive Summary, Underwriting Platform Inventory, Risk Assessment Model215Review, Pricing Adequacy Analysis, Portfolio Exposure Assessment, Workflow Automation216Maturity, Data Quality Scorecard, Model Governance Review, Prioritized Recommendations.217218219============================================================220SELF-HEALING VALIDATION (max 2 iterations)221============================================================222223After producing output, validate data quality and completeness:2242251. Verify all output sections have substantive content (not just headers).2262. Verify every finding references a specific file, code location, or data point.2273. Verify recommendations are actionable and evidence-based.2284. If the analysis consumed insufficient data (empty directories, missing configs),229 note data gaps and attempt alternative discovery methods.230231IF VALIDATION FAILS:232- Identify which sections are incomplete or lack evidence233- Re-analyze the deficient areas with expanded search patterns234- Repeat up to 2 iterations235236IF STILL INCOMPLETE after 2 iterations:237- Flag specific gaps in the output238- Note what data would be needed to complete the analysis239240============================================================241OUTPUT242============================================================243244## Underwriting Analysis Complete245246- Report: `docs/underwriting-analysis-report.md`247- Lines of business reviewed: [count]248- Risk models assessed: [count]249- Pricing gaps identified: [count]250- Compliance issues found: [count]251252### Summary Table253| Area | Status | Priority |254|------|--------|----------|255| Risk Classification | [PASS/WARN/FAIL] | [P1-P4] |256| Predictive Models | [PASS/WARN/FAIL] | [P1-P4] |257| Guidelines Engine | [PASS/WARN/FAIL] | [P1-P4] |258| Pricing Adequacy | [PASS/WARN/FAIL] | [P1-P4] |259| Portfolio Exposure | [PASS/WARN/FAIL] | [P1-P4] |260| Workflow Automation | [PASS/WARN/FAIL] | [P1-P4] |261| Data Quality | [PASS/WARN/FAIL] | [P1-P4] |262| Model Governance | [PASS/WARN/FAIL] | [P1-P4] |263264NEXT STEPS:265266- "Run `/actuarial-modeling` to evaluate loss reserving and premium pricing models."267- "Run `/catastrophe-modeling` to assess natural disaster exposure and PML estimates."268- "Run `/claims-workflow` to analyze claims adjudication and its impact on loss ratios."269270DO NOT:271272- Do NOT modify any rating algorithms, underwriting rules, or policy data.273- Do NOT access or display personally identifiable policyholder information.274- Do NOT make definitive rate adequacy conclusions without actuarial validation.275- Do NOT skip regulatory compliance checks even for surplus lines or non-admitted business.276- Do NOT assume predictive model fairness without testing for unfair discrimination.277278279============================================================280SELF-EVOLUTION TELEMETRY281============================================================282283After producing output, record execution metadata for the /evolve pipeline.284285Check if a project memory directory exists:286- Look for the project path in `~/.claude/projects/`287- If found, append to `skill-telemetry.md` in that memory directory288289Entry format:290```291### /underwriting-analysis — {{YYYY-MM-DD}}292- Outcome: {{SUCCESS | PARTIAL | FAILED}}293- Self-healed: {{yes — what was healed | no}}294- Iterations used: {{N}} / {{N max}}295- Bottleneck: {{phase that struggled or "none"}}296- Suggestion: {{one-line improvement idea for /evolve, or "none"}}297```298299Only log if the memory directory exists. Skip silently if not found.300Keep entries concise — /evolve will parse these for skill improvement signals.