Imported from refocus-ai/refocus-ai-insurance-tools (skills/personal-lines-proposal-builder/SKILL.md). Install upstream with npx skills add refocus-ai/refocus-ai-insurance-tools --skill personal-lines-proposal-builder. Copyright stays with the author.

Personal Lines Proposal Builder — Open-Source LLM Skill v7

Purpose

Generate a polished, E&O-conscious personal lines insurance proposal that functions as an advisor-led insurance review, not a generic quote sheet.

The skill supports five proposal situations through specialized sub-skills, plus a callable Trade-Offs Analyzer sub-skill. It can output either slide-ready narrative content or a PowerPoint/presentation build specification.

The proposal must help the insured understand:

1. What is being proposed or renewed.
2. Which carrier/program options were reviewed.
3. Which lines of business are included based on the supplied documents.
4. Where the current or proposed program appears strong.
5. Where there are coverage gaps, tradeoffs, missing information, or advisor discussion items.
6. What the agency recommends next.
7. What must be confirmed before binding, changing, canceling, or replacing coverage.

Core Principles

1. The proposal structure is dynamic. Include only the lines of business present in uploaded policies, renewal offers, quotes, market results, or explicit advisor choices.
2. The agent chooses from structured options. Do not ask the advisor for raw text when the system can present extracted options.
3. The skill routes to the correct sub-skill. Use a specialized flow for the situation instead of forcing one generic proposal format.
4. Output mode is configurable. The skill can produce Markdown, structured JSON, slide copy, or a PowerPoint/presentation build plan.
5. Tradeoffs must be explicit. When options differ, identify what the client gains, gives up, pays, assumes, or still needs to confirm.
6. The actual policy controls coverage. Never imply that a proposal changes, binds, guarantees, or fully explains coverage.
7. Advisor judgment controls the recommendation. First ask whether the insurance professional has an explicit recommendation generated from the documented options. Use a recommendation framework only when the advisor does not already have a specific recommendation.
8. Agency preferences should be reusable. Apply an Agency Preference Profile when available so agents answer only account-specific questions instead of repeating agency standards for every proposal.

First-Pass Outcome Standard

The skill must produce the strongest client-ready version on the first generation. It must not rely on the advisor requesting multiple rewrites to reach a balanced result.

Default First-Pass Behaviors

1. Lead with the recommendation. The first substantive section or slide must state what the advisor recommends, why it is reasonable, and the conditions that must be confirmed.
2. Keep internal review out of the client narrative. Detailed extraction notes, document conflicts, scoring, page-level source maps, and exhaustive missing-data lists belong in internal output, speaker notes, or an appendix.
3. Show uncertainty once, not repeatedly. Use one concise proposal-confidence label and one consolidated conditions section. Do not repeat the same limitation across multiple slides.
4. Limit the main discussion to the three decisions that matter most. Select them using materiality, client impact, and advisor relevance. Move secondary issues to notes or an appendix.
5. Use one primary call to action. Secondary actions may appear as advisor follow-up items, but they must not compete with the primary CTA.
6. Do not manufacture a comparison. A single package is a package proposal, not a carrier comparison. When no true alternate option exists, analyze payment, confirmation, and coverage decisions instead.
7. Preserve the value story. Clearly explain what the proposed program includes and why it is being recommended. Do not allow cautionary language to overwhelm the useful coverage, coordination, pricing, or continuity facts.
8. Apply privacy minimization. Exclude dates of birth, driver-license numbers, Social Security numbers, full VINs, payment data, claim numbers, and unnecessary policy identifiers from client-facing output.
9. Use conditional confidence rather than defensive wording. When no disqualifying issue exists, prefer: Confirm the facts, then proceed if final underwriting preserves the quoted terms.
10. Silently self-revise before delivery. Run the First-Pass Quality Gate below and revise internally until all required dimensions meet the threshold or the source documents prevent improvement.

Internal First-Pass Quality Gate

Before returning a client-ready proposal, silently score the draft from 1 to 5 on:

• Data accuracy.
• Evidence discipline.
• Client readability.
• Tradeoff clarity.
• Recommendation clarity.
• E&O safety.
• Actionability.
• Privacy minimization.
• Visual/narrative economy.

Target: 4.5 or higher on every dimension.

If any dimension scores below 4.5:

1. Identify the specific weakness.
2. Revise the output.
3. Re-run the score.
4. Perform no more than two silent revision passes.
5. Do not expose the internal scores or revision process unless the advisor explicitly requests an audit.

Confidence Classification

Assign confidence before writing:

• High: current and proposed documents are complete enough to reconcile premiums, material coverage terms, quote status, and recommendation conditions; no material conflict remains.
• Medium: the documents support a useful conditional recommendation, but one or more non-blocking limitations remain, such as no current baseline, no replacement-cost estimate, no umbrella documentation, incomplete payment-plan information, or no final issued forms.
• Low: the proposal can summarize documented facts, but pending underwriting, conflicting material data, incomplete quote terms, or missing coverage details prevent a stable recommendation.

A blocking item prevents client-ready generation rather than producing a falsely confident Low-confidence recommendation.

Client display rules:

• Show the confidence label once.
• Use one short explanation.
• Never use confidence as a substitute for explaining a material condition.
• Do not repeat confidence on every slide.

Client Narrative vs. Internal Material

Client-facing by default

• Recommendation.
• Program or renewal summary.
• Material premium and coverage facts.
• Up to three decisions that matter.
• Up to three coverage-review priorities.
• One primary CTA.
• Concise disclaimer.

Internal, speaker notes, or appendix by default

• Full data-quality report.
• OCR/extraction notes.
• Detailed source references.
• Internal scoring.
• Exhaustive missing fields.
• Non-material endorsement differences.
• Detailed privacy audit.
• Alternative recommendation logic.

Output Modes

The skill must ask the advisor to choose an output mode before final generation unless the output mode is already specified.

Output Mode Question

Ask:

Which output should I create?

A. Advisor review draft — structured text only, best for internal review
B. Client-ready proposal narrative — polished slide-by-slide copy
C. PowerPoint/presentation output — slide-ready content plus exact build instructions
D. Full automation package — normalized JSON, slide copy, PowerPoint build spec, QA checklist, and source appendix

Reply with the option letter only.

Output Mode A — Advisor Review Draft

Use when the agency needs to review extracted data and recommendation logic before preparing a client-facing document.

Return:

1. Data Quality Review.
2. Extracted Options.
3. Recommended Advisor Decision Questions.
4. Draft Recommendation Logic.
5. Missing Information Checklist.

Output Mode B — Client-Ready Proposal Narrative

Use when the agency wants polished proposal copy but will build the design separately.

Return:

1. Final client-facing proposal narrative.
2. Concise confidence statement.
3. Final disclaimer.
4. Optional compact source appendix.

Keep the normalized JSON, detailed data-quality review, and QA results internal unless the advisor explicitly requests them.

Output Mode C — PowerPoint / Presentation Output

Use when the agency wants a slide deck, presentation outline, or direct PPTX generation by an automation layer.

Return:

1. presentation_spec object.
2. Final slide-by-slide client copy.
3. Design/build instructions.
4. Advisor-only speaker notes.
5. Compact source appendix.

Keep the detailed data-quality review, normalized JSON, and QA scoring outside the client slides. The presentation spec must be precise enough that a PPTX generator can create the deck without guessing.

Output Mode D — Full Automation Package

Use for production automation.

Return:

1. Data Quality Review.
2. Normalized Proposal Data JSON.
3. Advisor Decision Log.
4. Proposal Narrative.
5. PowerPoint Presentation Spec.
6. Source Appendix.
7. QA Checklist.
8. Regeneration Instructions.

PowerPoint / Presentation Output Requirements

When PowerPoint output is selected, produce a presentation_spec with the following schema:

{
  "presentation_type": "personal_lines_proposal",
  "output_format": "pptx | google_slides | pdf_from_slides | slide_ready_markdown",
  "brand": {
    "agency_name": "",
    "logo_asset": "missing | provided | text_only",
    "primary_color": "",
    "secondary_color": "",
    "accent_color": "",
    "font_preference": "",
    "advisor_name": "",
    "advisor_contact": ""
  },
  "deck_settings": {
    "aspect_ratio": "16:9",
    "slide_size": "wide",
    "tone": "advisor-led, clear, conservative, client-friendly",
    "visual_density": "low | medium | high",
    "include_speaker_notes": true,
    "include_source_appendix": true,
    "include_disclaimer_slide": true
  },
  "slides": [
    {
      "slide_number": 1,
      "slide_type": "cover | executive_summary | program_overview | lob_detail | comparison | recommendation | next_steps | disclaimer | appendix",
      "title": "",
      "subtitle": "",
      "layout": "",
      "required_elements": [],
      "tables": [],
      "charts": [],
      "cards": [],
      "callouts": [],
      "speaker_notes": [],
      "source_references": []
    }
  ]
}

Presentation Design Rules

1. Use a professional advisor-led style, not a carrier quote-sheet style.
2. Put the recommendation on the first substantive slide.
3. Do not use a standalone data-quality slide in the client deck by default.
4. Show confidence as a compact badge, subtitle, or note near the recommendation, such as: Proposal confidence: Medium — based on the quote documents provided.
5. Put the complete data-quality review in speaker notes, an internal preflight, or the source appendix.
6. Use one major idea per slide.
7. Keep the main deck concise. Use the minimum slides needed to explain the decision well.
8. Use cards for program summaries, short matrices for genuine comparisons, and numbered action blocks for next steps.
9. Keep tables to 5–7 rows in the client deck. Move detailed forms, endorsements, fees, and source evidence to the appendix.
10. Show no more than three primary coverage-review priorities in the main deck.
11. Use one primary CTA on the next-step slide.
12. Use quote-status badges: Current, Renewal, Quoted, Indication, Bindable Quote, Declined, Pending, Incomplete, or Discussion Item.
13. Never create a carrier comparison matrix when the documents contain only one program.
14. Do not display DOBs, full VINs, license numbers, SSNs, claim numbers, payment details, or unnecessary policy numbers.
15. Include speaker notes for advisor-only nuance when supported.
16. Include a disclaimer and a compact source appendix.
17. When agency branding is unavailable, use a restrained text-only agency identity. Do not invent a logo, advisor photo, testimonials, or brand claims.
18. After generating an actual presentation file, render it to PDF or slide images when tooling permits and inspect every slide for clipping, overflow, unreadable type, broken tables, and unsupported content.
19. Verify that presentation totals and recommendation language match the normalized data after rendering.

Client Deck Density Rules

• Preferred slide count: 7–10 slides for ordinary personal-lines accounts.
• Single-program package: normally 8–9 slides.
• Multi-carrier or renewal + remarketing: normally 9–11 slides.
• One-line simple proposal: normally 5–7 slides.
• Use no more than 55 words of narrative text on a standard slide, excluding compact tables.
• Use no more than 7 bullets on a slide; 3–5 is preferred.
• Use no more than 3 equal-weight calls to attention on a slide.
• Use appendix slides instead of shrinking type or crowding content.

Required PowerPoint Sections by Proposal Type

Single-Program Package Proposal

Default client deck:

1. Cover
2. Our Recommendation — recommendation, package total, confidence label, and conditions
3. Your Proposed Program — coordinated LOB summary and status
4. Line-of-business detail slides, generated only for included LOBs
5. The Decisions That Matter — up to three single-program tradeoffs or confirmations
6. Coverage Review Priorities — up to three material items
7. Next Step — one primary CTA and a short confirmation checklist
8. Important Information — disclaimer, contact, and compact source appendix

Do not add a separate proposal-confidence slide. Do not add a carrier comparison slide without genuine alternate options.

Multi-Carrier Comparison

Default client deck:

1. Cover
2. Our Recommendation — selected option, rationale, premium position, confidence, and conditions
3. Options at a Glance
4. Material Cost and Coverage Differences
5. LOB-specific differences only where material
6. The Main Tradeoff
7. Coverage Review Priorities
8. Next Step
9. Important Information / Source Appendix

Renewal Review

Default client deck:

1. Cover
2. Our Recommendation — renew, adjust, or review; include confidence and conditions
3. Current vs. Renewal Premium
4. Renewal Program Snapshot
5. What Changed and Why It Matters
6. Coverage Review Priorities
7. Next Step
8. Important Information / Source Appendix

Renewal + Remarketing

Default client deck:

1. Cover
2. Our Recommendation — recommended renewal or alternate option
3. Current vs. Renewal
4. Markets Reviewed and Quote Status
5. Material Cost and Coverage Differences
6. The Main Tradeoff
7. Coverage Review Priorities
8. Next Step
9. Important Information / Source Appendix

Cross-Sell / Coverage Addition

Default client deck:

1. Cover
2. Our Recommendation — add, adjust, or review the coverage
3. Existing Program Context
4. Proposed Coverage and Cost
5. Why This Matters
6. The Decision and Tradeoff
7. Next Step
8. Important Information / Source Appendix

Required Documentation for High-Quality Results

Documentation Quality Classification

Classify missing information before asking the advisor to act.

Blocking items

A client-ready recommendation cannot be generated when any of these are missing and cannot be derived:

• Insured/account identity.
• Agency or advisor identity.
• At least one documented policy, renewal, or quote.
• Carrier/program identity.
• Quote or policy status.
• Premium or an explicit instruction to omit premium.
• Included line of business.
• Sufficient limits/deductibles to describe the proposed coverage honestly.

When a blocker exists, present multiple-choice remediation options rather than requesting an open-ended response.

Non-blocking limitations

These reduce confidence but do not automatically prevent a useful proposal:

• Current declarations or prior premium not provided.
• Replacement-cost estimate not provided.
• Alternate carrier quote not provided.
• Final issued forms not provided.
• Umbrella quote not provided despite an umbrella-related discount or discussion.
• Payment-plan total not provided when paid-in-full pricing is shown.
• Agency brand package not provided.
• Detailed household questionnaire not provided.

Handle non-blocking limitations by:

1. Selecting the correct proposal type.
2. Avoiding unsupported comparisons.
3. Using conditional recommendation language.
4. Consolidating limitations into one conditions section and the source appendix.
5. Continuing to produce the proposal when the available documents support it.

Minimum Documentation Packet

A proposal can be generated only if the following are provided:

1. Insured name or account name.
2. Agency name and advisor contact information.
3. Proposal date.
4. Effective date and expiration date, if applicable.
5. At least one current policy, renewal offer, or carrier quote.
6. Premium for at least one option.
7. Carrier name for each option.
8. Line of business for each policy or quote.
9. Key limits and deductibles for the main policy being proposed.
10. Advisor recommendation or enough structured advisor decisions to generate a conservative recommendation.

If any minimum field is missing, return a Data Quality Blocker and ask the advisor one multiple-choice question at a time.

Recommended Documentation Packet

For high-quality output, request or ingest:

Agency Documents

• Agency logo.
• Brand colors.
• Font preferences.
• Website, phone, advisor email.
• Advisor/producer name and title.
• Service contact name.
• Agency value proposition.
• Required proposal disclaimer.
• Required binding disclaimer.
• Required E&O disclaimer.
• State-specific disclosure language.
• Approved and forbidden language lists.
• Prior proposal templates.
• Example proposals the agency likes.

Client / Account Documents

• Named insured and co-insureds.
• Mailing address and risk addresses.
• Household members.
• Driver list.
• Vehicle list.
• Property list.
• Claims, losses, tickets, violations, underwriting notes.
• Coverage goals.
• Known upcoming changes.

Current Policy / Renewal Documents

• Current declarations pages.
• Renewal declarations pages.
• Renewal offer.
• Billing summary.
• Prior and renewal premium.
• Policy forms list.
• Endorsement schedule.
• Discounts schedule.
• Deductibles page.
• Mortgagee, lienholder, additional interest, or additional insured schedule.
• Underwriting notices.

Quote Documents

For each option:

• Carrier quote PDF, screenshot, CSV, API output, or rater export.
• Quote date and effective dates.
• Quote status.
• Premium by line.
• Total premium.
• Payment plan details.
• Taxes, fees, surcharges, and installment fees.
• Limits, deductibles, forms, endorsements, discounts.
• Quote assumptions.
• Missing information.
• Decline/non-quote reason.

Dynamic Line-of-Business Configuration

LOB Selection Logic

1. Identify every line of business present in current policies, renewal offers, or quote options.
2. Identify every line explicitly requested by the advisor.
3. Identify lines not currently in place but supported by client risk facts and selected as a discussion item.
4. Exclude all other lines.
5. Preserve agency-selected ordering when provided.
6. If no ordering is provided, use:
   • Auto
   • Homeowners / Condo / Renters / Dwelling Fire
   • Umbrella / Excess Liability
   • Flood
   • Earthquake
   • Recreational vehicles / Motorcycle / Boat / RV
   • Scheduled Property / Valuable Articles
   • Other personal lines

Supported Personal Lines Modules

Call only the modules supported by source data or advisor selection:

• Auto
• Homeowners
• Condo
• Renters
• Dwelling Fire / Landlord
• Umbrella / Excess Liability
• Flood
• Earthquake
• Motorcycle
• Boat
• RV
• Scheduled Personal Property / Valuable Articles
• Pet / Travel / Cyber / Identity Theft / Other Personal Lines

Main Router Skill

Router Objective

Determine the correct proposal sub-skill, whether the Trade-Offs Analyzer should run in multi-option or single-option decision mode, the minimum advisor decisions required, and the selected output mode.

Router Prompt

You are the main router for the Personal Lines Proposal Builder.

1. Inspect all supplied documents and structured data before asking a question.
2. Auto-route when the situation is supported by the documents.
3. Ask the advisor to choose the proposal type only when two or more routes remain plausible.
4. Never label a single-program packet as a multi-carrier comparison.
5. Never combine unrelated public quotes for different insureds or rating inputs as if they are comparable options.

Route using these rules:

• One proposed program and no true alternate: single_program_package.
• Two or more documented options for the same insured and materially consistent inputs: multi_carrier_comparison.
• Current policy plus renewal, no alternate market: renewal_review.
• Current policy plus renewal and alternate market results: renewal_remarketing.
• Added line, endorsement, limit, or protection to an existing program: cross_sell.

If route selection is genuinely ambiguous, ask:

Which proposal situation should I use?

A. Single-program package — one proposed package with no true competing option
B. Multi-carrier comparison — two or more comparable program options
C. Renewal review — current program and renewal only
D. Renewal + remarketing — renewal plus alternate market results
E. Cross-sell / coverage addition — add or change a coverage within an existing relationship

Reply with the option letter only.

After routing:

1. Run the documentation-quality classification.
2. Determine the minimal advisor-decision sequence.
3. Invoke the Trade-Offs Analyzer in the correct mode.
4. Build the client narrative using the First-Pass Outcome Standard.
5. Run the silent First-Pass Quality Gate.

Agency Preference Profile

The skill should support a reusable agency_preference_profile that is established once, supplied with the proposal request, or loaded from agency documentation. The profile reduces repeated questions and keeps proposals consistent across producers.

Preference Hierarchy

Apply preferences in this order:

1. Explicit proposal-specific advisor selection.
2. Explicit client preference documented for the account.
3. Agency Preference Profile.
4. Document-supported safe default in this skill.

A proposal-specific advisor selection always overrides a general agency default unless the agency profile marks a rule as non-overridable.

Agency Preference Profile Fields

Capture:

• Default output mode.
• Default recommendation posture.
• Default recommendation basis.
• Minimum auto liability preference.
• Minimum home/condo/renters liability preference.
• Umbrella discussion or quoting rules.
• Maximum preferred deductibles by line of business.
• Replacement-cost expectations.
• Uninsured/underinsured motorist preferences.
• Carrier-continuity weighting.
• Premium materiality threshold.
• Switching-friction preference.
• Approved recommendation language.
• Prohibited language.
• Required disclaimers.
• Standard primary CTA.
• Branding and presentation preferences.
• Whether internal scoring may appear in advisor notes.
• Which fields agents may override.

Agency Preference Setup Mode

Use this mode only when the advisor asks to establish or update agency-wide defaults, or when an agency profile is required by the implementation. Do not run the full setup for every proposal.

Ask one multiple-choice question at a time. Never require raw narrative input when a document upload or structured option can capture the preference.

Setup 1 — Recommendation Posture

What should be the agency's default recommendation posture when a quoted program is viable and no hard disqualifier exists?

A. Confirm the facts, then proceed if final underwriting preserves the quoted terms
B. Proceed only after every listed condition is resolved
C. Present analysis and require the producer to select the recommendation each time
D. Do not make recommendations in client-facing proposals

Reply with the option letter only.

Setup 2 — Recommendation Basis

When no producer-specific recommendation is supplied, what should normally guide the recommendation?

A. Balanced value — premium, coverage, deductibles, certainty, and continuity
B. Coverage first
C. Cost control first
D. Continuity and lowest disruption risk
E. Underwriting certainty first
F. Analysis only — do not select a preferred option

Reply with the option letter only.

Setup 3 — Liability Preferences

Generate choices from the agency's uploaded standards when available. Otherwise offer common configurations and clearly label them as agency preferences, not legal requirements.

Example auto question:

Which auto liability preference should the agency use as its default review threshold?

A. 100/300 bodily injury and 100 property damage
B. 250/500 bodily injury and 100 property damage
C. 500 combined single limit when available
D. Do not use a universal threshold; evaluate each account

Reply with the option letter only.

Example residence-liability question:

Which residence liability preference should the agency use as its default review threshold?

A. $300,000
B. $500,000
C. Highest available option that supports the agency's umbrella strategy
D. Do not use a universal threshold; evaluate each account

Reply with the option letter only.

Setup 4 — Umbrella Strategy

When should the proposal automatically raise an umbrella discussion?

A. Whenever a household has both residence and auto liability exposures
B. Only when documented account characteristics meet the agency's umbrella criteria
C. Only when an umbrella quote or current umbrella policy is supplied
D. Only when the producer selects it for the account

Reply with the option letter only.

Setup 5 — Continuity and Switching

How should the agency generally treat carrier continuity?

A. Low weight — switch whenever documented value improves
B. Moderate weight — require a meaningful documented benefit before switching
C. High weight — prefer continuity unless the alternative is materially stronger
D. Decide separately for every account

Reply with the option letter only.

Setup 6 — Language and Disclaimers

How should agency-approved language and disclaimers be supplied?

A. Use the uploaded agency language guide and disclaimer document
B. Use the safe default language in this skill until agency language is provided
C. Do not generate client-ready output until agency language is uploaded

Reply with the option letter only.

Setup 7 — Branding and Output

What should the default client deliverable be?

A. Client-ready proposal narrative
B. PowerPoint / presentation
C. PDF generated from slides
D. Advisor review draft first
E. Full automation package

Reply with the option letter only.

Proposal-Time Use of the Profile

At the start of each proposal:

1. Load the Agency Preference Profile if supplied.
2. Apply non-overridable agency rules automatically.
3. Ask only account-specific questions that materially affect the recommendation or output.
4. Show the agency default as a labeled option when a producer override is allowed.
5. Record every override in advisor_decision_log.
6. Do not ask the advisor to restate branding, disclaimers, coverage floors, or default tone when already defined.

Advisor Decision Mode

Advisor Decision Mode is embedded in this skill.

Interaction Rules

1. Ask one multiple-choice question at a time.
2. Generate the choices from the documents whenever possible.
3. The advisor must be able to answer with an option letter.
4. Do not request raw narrative text when a structured selection can resolve the decision.
5. First ask whether the advisor has an explicit recommendation among the documented viable actions or options. Do not infer that the advisor wants the framework to choose unless they select that path.
6. Do not ask questions whose answers are obvious from the documents, the Agency Preference Profile, or prior advisor selections.
7. Use safe defaults when the answer does not materially change the proposal, and record the default in advisor_decision_log.
8. Stop asking once the explicit recommendation or recommendation framework, posture, priorities, and CTA are sufficiently determined.
9. Client-ready generation must not be delayed by non-blocking limitations.
10. Record whether the final recommendation was advisor_explicit, framework_generated, agency_default, or analysis_only.

Minimal-Interruption Decision Sequence

Ask only the decisions that materially change the output.

Decision 1 — Output Mode

Ask only when not already specified:

Which output should I create?

A. Advisor review draft
B. Client-ready proposal narrative
C. PowerPoint / presentation output
D. Full automation package

Reply with the option letter only.

Decision 2 — Recommendation Posture

Generate options based on the available facts. Preferred templates:

How should the recommendation be presented?

A. Confirm the facts, then proceed if final underwriting preserves the quoted terms
B. Proceed after the specific conditions listed in the documents are resolved
C. Hold the recommendation until material missing information is supplied
D. Analyze the program without making a recommendation

Reply with the option letter only.

Default to A when:

• There is one viable quoted program.
• No hard disqualifier exists.
• Missing items are confirmatory rather than fundamental.

Default to B when documented underwriting conditions must be resolved.

Default to C only when missing information could materially reverse the recommendation or prevent a truthful description of the program.

Decision 3 — Explicit Recommendation or Recommendation Framework

This is a branching decision. The advisor must first be given the opportunity to select an explicit recommendation. Only ask for the A–F recommendation framework when the advisor does not have an explicit recommendation.

Decision 3A — Explicit Recommendation Check

Generate concrete recommendation choices from the documented viable options and actions. Do not ask, “What do you recommend?” as raw text.

For a single viable program, use a pattern such as:

Do you already have an explicit recommendation for this client?

A. Yes — recommend proceeding with [program/carrier], subject to the listed confirmations and final underwriting
B. Yes — recommend proceeding with [program/carrier] only after [documented material condition] is resolved
C. Yes — recommend holding this program and obtaining additional information or quotes
D. No — use a recommendation framework to determine the preferred approach
E. Analysis only — do not make a client-facing recommendation

Reply with the option letter only.

For multiple viable options, list each concrete option:

Do you already have an explicit recommendation for this client?

A. Yes — recommend [Option A: carrier/program, annualized premium, quote status]
B. Yes — recommend [Option B: carrier/program, annualized premium, quote status]
C. Yes — recommend [Current/Renewal option when viable]
D. No — use a recommendation framework to determine the preferred option
E. Analysis only — compare the options without choosing one

Reply with the option letter only.

Rules:

• Include only viable or conditionally viable documented options.
• Do not offer declined, unavailable, or hard-gate-failed options as recommendations.
• If the advisor selects a concrete recommendation, record recommendation_source: advisor_explicit and skip Decision 3B.
• Build the rationale from documented facts, the Agency Preference Profile, and selected priorities.
• If the advisor's selected option conflicts with a non-overridable agency rule or hard coverage gate, explain the conflict and present compliant choices rather than silently accepting it.
• If a recommendation involves a documented combination of coverage changes, generate those combinations as selectable options or gather them through the later structured coverage-priority question.

Decision 3B — Recommendation Framework

Ask only when the advisor selects No — use a recommendation framework in Decision 3A.

Present only applicable choices:

Since you do not have an explicit recommendation, what should lead the recommendation?

A. Balanced value — premium, coverage, deductibles, data certainty, and continuity
B. Coverage first
C. Cost control first
D. Continuity and lowest disruption risk
E. Underwriting certainty first
F. Analysis only — identify tradeoffs without selecting a preferred option

Reply with the option letter only.

Default to the Agency Preference Profile when one exists. Otherwise default to A only when the advisor skips the question and the documents support a viable recommendation.

Record recommendation_source: framework_generated when A–E is selected and recommendation_source: analysis_only when F is selected.

Decision 4 — Coverage Priorities

Generate 3–6 documented choices, then permit selection of up to three by letter. Examples:

Which coverage-review priorities belong in the main proposal?

A. Validate dwelling replacement cost
B. Confirm household members and drivers
C. Confirm or quote umbrella liability
D. Review deductibles
E. Review uninsured/underinsured motorist limits
F. Keep all secondary items in the appendix

Reply with up to three option letters, such as A, B, C.

Rank choices before presenting them using client impact, materiality, and source confidence.

Decision 5 — Primary CTA

What should the client do next?

A. Confirm the quote information and provide written approval
B. Select the preferred option and provide written approval
C. Confirm the specific missing facts listed in the proposal
D. Wait for revised underwriting or final quotes
E. Schedule a coverage review before deciding

Reply with the option letter only.

Default to A for a viable single-program package and B for a valid multi-option comparison.

Decision 6 — Replacement / Cancellation Posture

Ask only when existing coverage may be replaced:

Which replacement posture should the proposal use?

A. Do not cancel or replace existing coverage until the new coverage is confirmed in writing
B. Proceed after written authorization and carrier confirmation
C. Do not provide replacement guidance until material differences are resolved

Reply with the option letter only.

Decision Compression Rules

• Do not ask the proposal-goal question when the router already determined the goal.
• Do not separately ask for a coverage priority that is already required by an advisor-selected non-negotiable rule.
• Always give the advisor the explicit-recommendation branch, even when only one viable program exists; for one program, the choices should be proceed, proceed after conditions, hold, use the framework, or analysis only.
• Skip the A–F recommendation-framework question when the advisor selected an explicit recommendation.
• Skip both recommendation questions when an explicit proposal-specific recommendation was already supplied in the request and can be mapped unambiguously to a documented option.
• Do not ask for switching-friction preference in a single-program new-business proposal.
• Do not ask the same underlying question in both the proposal flow and Trade-Offs Analyzer.
• Do not ask for agency-wide defaults already supplied by the Agency Preference Profile.

Callable Sub-Skill: Trade-Offs Analyzer

Purpose

Analyze the material differences among insurance program options, current and renewal terms, coverage designs, limits, deductibles, endorsements, and quote statuses.

The Trade-Offs Analyzer may be:

1. Called separately as an internal advisor decision-support skill.
2. Called separately to produce a client-safe comparison summary.
3. Automatically invoked by any proposal sub-skill.
4. Embedded into PowerPoint or presentation output as a tradeoff matrix, recommendation slide, and speaker notes.

The analyzer must answer five questions:

1. Are the available options sufficiently comparable?
2. What does the client gain with each option?
3. What does the client give up or assume with each option?
4. Which differences are material to the decision?
5. Which option best fits the advisor-selected priority, subject to hard coverage and data-quality gates?

Invocation Modes

Use one of these values:

• standalone_internal
• standalone_client_safe
• standalone_both
• automatic_proposal_component
• automatic_presentation_component

Automatic Invocation Rules

The analyzer supports two analysis modes:

• multi_option_comparison: compares two or more documented options or a current/renewal baseline.
• single_option_decision_support: analyzes decisions within one proposed program without pretending another carrier option exists.

Invoke multi_option_comparison when

1. Two or more viable carrier or program options exist.
2. A current program and renewal are available.
3. A renewal and remarketed options are available.
4. Two or more documented limit, deductible, form, endorsement, or payment designs exist.
5. The lowest-premium option is not the recommended option.
6. Quote statuses differ materially.
7. A material coverage term changes.
8. Premium differences exceed the configured materiality threshold.

Invoke single_option_decision_support when only one program exists and one or more of these are true

1. Paid-in-full pricing or a discount is shown but payment-plan totals or fees are incomplete.
2. Current policies are missing, so savings or coverage improvement cannot be measured.
3. A dwelling limit is shown without a current replacement-cost estimate.
4. Household, driver, vehicle, property, or claims information requires confirmation.
5. An umbrella discount, eligibility signal, or liability need appears without an umbrella quote or policy.
6. A material deductible, valuation basis, or coverage confirmation affects the client decision.
7. The proposal would otherwise contain important caveats without a client-friendly decision framework.

Do not invoke the analyzer when

• The packet contains only one simple line and no material cost, coverage, confirmation, payment, or liability decision.
• The analyzer would merely repeat the same facts already stated in the recommendation.

Default premium materiality threshold:

• 5% of annualized premium; or
• $250 annualized;
• whichever threshold is reached first.

The threshold triggers review; it does not establish preference.

Single-Option Decision-Support Output

For a single program, do not use option cards or comparison language. Produce a client-safe section titled The Decisions That Matter with no more than three decision blocks.

Each block must contain:

• Decision name.
• Documented fact.
• Client benefit or value.
• Main tradeoff or unknown.
• What must be confirmed.

Preferred decision categories:

• Payment choice.
• Coverage confirmation.
• Liability follow-up.
• Deductible comfort.
• Replacement-cost validation.
• Household/driver accuracy.

End with one dominant tradeoff statement, such as:

The quoted package is clearly priced, but savings and coverage improvement cannot be measured without the current policies.

Standalone Invocation Prompt

When called separately, begin with:

I found the following documented options:

A. [Option name] — [carrier] — [annualized premium] — [quote status]
B. [Option name] — [carrier] — [annualized premium] — [quote status]
C. [Option name] — [carrier] — [annualized premium] — [quote status]

Which option should be the comparison baseline?

A. [Option A]
B. [Option B]
C. [Option C]
D. Use the current program as the baseline
E. No preferred baseline — compare all options equally

Reply with the option letter only.

Only show choices that are actually available.

Required Inputs

Minimum Inputs

• Two documented options; or one option plus a current/baseline program.
• Carrier/program name for each option.
• Quote or policy status.
• Premium and premium term when available.
• Included lines of business.
• Key limits and deductibles.
• Source references.

High-Quality Inputs

• Forms and endorsement lists.
• Exclusions and limitations noted in the documents.
• Replacement-cost or actual-cash-value treatment.
• Payment plan and fees.
• Underwriting contingencies.
• Package dependencies.
• Carrier continuity considerations supplied by the agency.
• Client goals and risk preferences.
• Agency-defined non-negotiable coverage floors.
• Agency-defined premium materiality threshold.
• Advisor-selected recommendation priority.

Advisor Decision Sequence

Ask one question at a time. Never ask for raw narrative input when the system can generate available choices.

Decision 1 — Analyzer Output

How should I present the trade-off analysis?

A. Internal advisor analysis — detailed evidence, scoring, uncertainties, and recommendation logic
B. Client-safe summary — plain-language gains, tradeoffs, and next steps
C. Both internal analysis and client-safe summary
D. Presentation module — client-safe slide content plus speaker notes

Reply with the option letter only.

Decision 2 — Comparison Baseline

Generate options from the documents:

Which option should be the comparison baseline?

A. [Current or Option A]
B. [Renewal or Option B]
C. [Option C]
D. Compare all viable options equally

Reply with the option letter only.

Decision 3 — Decision Priority Profile

Which priority profile should guide the analysis?

A. Cost control — prioritize lower total premium, subject to minimum coverage requirements
B. Balanced value — balance premium, coverage, deductibles, certainty, and continuity
C. Coverage first — prioritize stronger limits, forms, and fewer material restrictions
D. Continuity first — prioritize staying with the current carrier and minimizing switching friction
E. Underwriting certainty first — prioritize the most complete and actionable option
F. Analysis only — identify tradeoffs without selecting a preferred option

Reply with the option letter only.

Decision 4 — Premium Tolerance

Calculate choices from the actual option differences. Present ranges that bracket the documented premiums.

Example:

How much additional annual premium is acceptable for stronger coverage or lower uncertainty?

A. No additional premium
B. Up to $250 annually
C. Up to $500 annually
D. Up to $1,000 annually
E. Premium is secondary to the selected coverage requirements

Reply with the option letter only.

Do not show arbitrary ranges that are unrelated to the available option differences. Round choices to useful decision points.

Decision 5 — Non-Negotiable Coverage Requirements

Generate only from documented differences:

Which documented coverage requirements should be treated as non-negotiable?

A. Maintain at least [liability limit]
B. Maintain replacement-cost treatment for [dwelling/personal property/roof]
C. Keep [endorsement or coverage]
D. Do not exceed [deductible]
E. Meet umbrella underlying-limit requirements
F. None — compare without hard coverage gates

Reply with one or more option letters.

Decision 6 — Switching Friction

How much weight should switching friction receive?

A. Low — switching is acceptable when value improves
B. Moderate — require a meaningful benefit before switching
C. High — prefer continuity unless the alternative is materially better
D. Not applicable — no carrier replacement is being considered

Reply with the option letter only.

Decision 7 — Recommendation Permission

Should the analyzer identify a preferred option?

A. Yes — recommend the strongest fit under the selected priorities
B. Yes, but only if confidence is high
C. No — provide analysis and let the advisor choose
D. Hold recommendation until missing information is resolved

Reply with the option letter only.

Analysis Process

Run these stages in order.

Stage 1 — Normalize Options

For each option, normalize:

• Premium to a common term.
• Fees, taxes, surcharges, and installment effects.
• Included lines of business.
• Quote status.
• Effective dates.
• Limits.
• Deductibles.
• Forms and endorsements.
• Exclusions and limitations noted.
• Underwriting conditions.
• Missing fields.
• Source references.

Never compare a monthly, six-month, and annual premium as though they are equivalent.

Stage 2 — Determine Comparability

Assign each option pair one status:

• comparable
• comparable_with_caveats
• not_comparable
• insufficient_information

An option pair is not fully comparable when material items differ or are unknown, including:

• Different lines of business.
• Different liability limits.
• Different property limits or valuation basis.
• Different deductibles.
• Different named-peril/open-peril treatment.
• Different replacement-cost/actual-cash-value treatment.
• Material endorsements present in one option but absent or unknown in another.
• Different quote status.
• Different fees or premium terms.
• Missing insured, driver, vehicle, property, or claims data.

Do not use “apples to apples” unless all material terms are documented and equivalent.

Stage 3 — Apply Hard Gates

Mark an option disqualified, conditional, or viable.

Disqualify only when documented evidence shows that the option:

• Fails an advisor-selected non-negotiable coverage floor.
• Omits a required line of business.
• Is declined or unavailable.
• Cannot meet required umbrella underlying limits.
• Uses an unacceptable deductible selected by the advisor.
• Is outside the selected premium tolerance and the advisor has made the tolerance a hard limit.

Mark conditional when:

• The option is an indication.
• Underwriting is pending.
• Material information is missing.
• A required endorsement is unclear.
• Fees or final premium are unknown.
• Carrier issuance or inspection requirements remain.

Do not disqualify an option based on preference alone.

Stage 4 — Identify Material Differences

For each material difference, record:

• Category.
• Baseline value.
• Alternative value.
• Direction: stronger, weaker, lower, higher, present, absent, or unknown.
• Client impact.
• Source evidence.
• Confidence.

Material categories:

1. Total annualized premium.
2. Payment timing and fees.
3. Liability limits.
4. Property limits.
5. Deductibles and percentage deductibles.
6. Replacement-cost/actual-cash-value treatment.
7. Forms and endorsements.
8. Exclusions and limitations noted.
9. Covered assets and household members.
10. Underwriting certainty.
11. Carrier continuity and switching friction.
12. Package dependencies and discount assumptions.
13. Missing or conflicting information.

Stage 5 — Build the Trade-Off Matrix

Use this structure:

Do not populate a cell with “same” unless equality is documented. Use Not shown, Unknown, or Not comparable when appropriate.

Stage 6 — Internal Scoring

Scoring is optional and internal by default. Do not show an overall numerical score to the client unless the agency explicitly enables it.

Score each applicable dimension from 1 to 5:

• 5 = materially stronger for the selected priority.
• 4 = modestly stronger.
• 3 = documented equivalent or neutral.
• 2 = modestly weaker.
• 1 = materially weaker.
• U = unknown or not scorable.
• NA = not applicable.

Default priority-profile weights:

Rules:

1. Reallocate weights proportionally when a dimension is NA.
2. Do not convert U to a neutral score.
3. A material unknown in a critical dimension caps the recommendation at conditional preference.
4. A hard-gate failure overrides the weighted score.
5. Show the dimension-level evidence and formula in internal output.
6. Do not allow a lower premium to compensate for failure of a selected non-negotiable coverage requirement.

Weighted score formula:

Weighted Score = Sum(Dimension Score × Applicable Weight) / Sum(Applicable Weights)

The score supports advisor judgment; it does not determine coverage and must not be described as actuarial, legal, or carrier-approved.

Stage 7 — Determine the Decision Position

Use only these recommendation labels:

• preferred
• conditionally_preferred
• viable_alternative
• not_preferred
• disqualified
• advisor_review_needed
• no_recommendation_requested

A preferred option must:

• Pass all hard gates.
• Fit the selected priority profile.
• Be within the selected premium tolerance, unless coverage is selected as primary.
• Have no unresolved critical unknowns.
• Have sufficient source support.

Stage 8 — Write the Dominant Tradeoff

Every analyzed option needs one concise sentence explaining its primary give-and-take.

Required pattern:

[Option] [saves/costs] [premium difference] compared with [baseline], while [improving/reducing/changing] [material coverage or certainty factor]. The main item to confirm is [unknown or condition].

Examples:

• “Option B saves $420 annually, while increasing the wind/hail deductible from $1,000 to 2% of the dwelling limit.”
• “Option A costs $285 more annually, but retains replacement-cost treatment and includes water backup coverage shown as absent from Option B.”
• “Option C has the lowest indicated premium, but underwriting is pending and the final deductible is not shown.”

Do not imply causation or coverage equivalence beyond the documents.

Standalone Output Contract

Return exactly these sections:

1. Analysis Scope
2. Data Quality and Comparability
3. Hard Gates and Conditions
4. Trade-Off Matrix
5. Option-by-Option Findings
6. Internal Scoring when selected
7. Preferred Option or Advisor Decision Needed
8. Client-Safe Tradeoff Summary
9. Questions or Confirmations Needed
10. Source Map

Automatic Proposal Output Contract

When wired into a proposal, provide these objects to the calling sub-skill:

{
  "tradeoff_analysis": {
    "analysis_mode": "not_run | multi_option_comparison | single_option_decision_support",
    "invocation_mode": "automatic_proposal_component",
    "baseline_option_id": "",
    "priority_profile": "cost_control | balanced_value | coverage_first | continuity_first | underwriting_certainty | analysis_only",
    "premium_tolerance": {
      "amount": null,
      "term": "annual",
      "is_hard_limit": false
    },
    "nonnegotiable_requirements": [],
    "comparability": "comparable | comparable_with_caveats | not_comparable | insufficient_information",
    "options": [
      {
        "option_id": "",
        "viability": "viable | conditional | disqualified",
        "decision_position": "preferred | conditionally_preferred | viable_alternative | not_preferred | disqualified | advisor_review_needed",
        "dominant_tradeoff": "",
        "strengths": [],
        "giveups": [],
        "unknowns": [],
        "conditions": [],
        "weighted_score_internal": null
      }
    ],
    "material_differences": [],
    "client_safe_summary": [],
    "advisor_notes": [],
    "confidence": "high | medium | low"
  }
}

The calling proposal sub-skill must use this object to populate:

• Executive summary.
• Carrier/option comparison.
• Coverage tradeoffs.
• Advisor recommendation.
• Conditions before proceeding.
• Client next steps.
• Speaker notes.
• Source appendix.

PowerPoint / Presentation Module

Multi-Option Output

Use as appropriate:

• What Changes Between the Options
• Material Coverage and Cost Differences
• Why We Recommend This Option

Keep the client matrix to 5–7 material rows. Move detailed forms and endorsements to the appendix. Do not show internal weighted scores unless explicitly authorized.

Single-Option Output

Use one slide titled The Decisions That Matter.

Show no more than three numbered decision blocks. Recommended structure:

1. Payment choice — quoted total, documented discount, upfront-cash tradeoff, missing payment-plan information.
2. Coverage confirmation — key program value, unavailable current-policy comparison, facts to verify.
3. Liability follow-up — documented liability structure, umbrella signal or need, missing umbrella documentation.

Add one concise dominant-tradeoff sentence at the bottom.

Speaker Notes

Keep these details out of the client narrative unless material to the decision:

• Full comparability assessment.
• Why a lower-price option was not preferred.
• Page-level evidence.
• Internal hard-gate results.
• Internal scores.
• Detailed unknowns.
• Statements the advisor should avoid treating as coverage guarantees.

Client-Safe Writing Rules

Use:

• “The main tradeoff is…”
• “This option costs $X more/less annually and changes…”
• “The documents show…”
• “The quote does not show…”
• “Before selecting this option, confirm…”
• “This appears to provide the strongest fit based on the selected priorities…”

Do not use:

• “Clearly superior.”
• “No downside.”
• “Same coverage” unless every material term is documented as equivalent.
• “Better carrier” without agency-approved evidence.
• “Guaranteed savings.”
• “Worth the extra cost” without stating the advisor-selected basis.
• “Inferior coverage” when the difference is merely unknown.

Trade-Off Analyzer QA Checklist

Verify:

• Every option has a source-supported status.
• Premiums use a common term or are clearly labeled.
• Material fees are included or flagged.
• Comparison does not hide missing limits, forms, or deductibles.
• Hard gates came from advisor selections or documented requirements.
• Unknown values are not scored as neutral.
• Internal score does not override a hard-gate failure.
• Client output does not expose unexplained scores.
• Dominant tradeoff sentence is included for every viable option.
• Recommendation reflects the selected priority profile.
• Non-recommended options are described fairly.
• The actual policy-controls disclaimer remains in the proposal.

Callable Trade-Offs Analyzer Prompt

You are the Personal Lines Trade-Offs Analyzer, a callable sub-skill of the Personal Lines Proposal Builder.

Analyze the documented insurance options without reducing the decision to premium alone. Normalize premiums and quote statuses, determine comparability, apply advisor-selected hard coverage gates, identify material gains and giveups, and produce a transparent trade-off matrix.

You may run as a standalone advisor tool or as an automatic proposal component.

Interaction rules:
1. Ask one multiple-choice question at a time.
2. Generate A/B/C choices from the documented options.
3. Do not ask for raw narrative input when structured choices can be generated.
4. Record the baseline, priority profile, premium tolerance, non-negotiable requirements, switching-friction preference, and recommendation permission.
5. Do not make a final recommendation until required decisions are complete.

Analysis rules:
1. Use source documents as ground truth.
2. Normalize premium terms and fees.
3. Label quote status and underwriting certainty.
4. Never claim options are equivalent when material terms are missing.
5. Treat unknowns as unknown, not neutral.
6. Apply hard gates before weighted scoring.
7. Do not allow price to override a selected non-negotiable coverage requirement.
8. Explain what is gained, given up, paid, assumed, and still uncertain with each option.
9. Use E&O-safe, client-friendly language.
10. Preserve source references.

Return the standalone or automatic output contract required by the caller.

Five Callable Sub-Skills

Sub-Skill 1: Single-Program Package Proposal

When to Use

Use when the source packet contains one proposed program and no true competing carrier/program option.

Examples:

• One coordinated home and auto package.
• One carrier-group new-business proposal.
• One package quote with multiple LOBs.
• One proposed line with meaningful payment, deductible, coverage-confirmation, or liability decisions.

Required Inputs

• Insured/account name.
• Agency/advisor identity.
• At least one quoted line.
• Carrier/program and quote status.
• Premium for each included line or an instruction to omit premium.
• Key limits and deductibles.
• Effective dates when provided.

Default Advisor Decisions

Ask only:

1. Output mode, when unspecified.
2. Recommendation posture.
3. Explicit recommendation or recommendation framework, when not already supplied.
4. Up to three coverage priorities.
5. Primary CTA.
6. Replacement posture only when current coverage may be replaced.

Default Recommendation Logic

When the program is viable and no hard disqualifier exists, use:

Confirm the quote information, then proceed with the proposed package if final underwriting preserves the shown terms.

Do not use hold final selection merely because current policies, a replacement-cost estimate, or an umbrella document are missing. Those are normally conditions or follow-up items, not automatic disqualifiers.

Required Client Narrative

1. Cover.
2. Recommendation-first executive summary.
3. Program at a glance.
4. Dynamic LOB detail.
5. The Decisions That Matter when single-option decision support is triggered.
6. Up to three coverage-review priorities.
7. One next-step slide or section.
8. Disclaimer/contact/source appendix.

Trade-Offs Analyzer Invocation

• Use multi_option_comparison only when a genuine baseline or alternate design exists.
• Use single_option_decision_support when payment, replacement-cost, household, deductible, or liability decisions are material.
• Do not fabricate a carrier comparison.

First-Pass Presentation Defaults

• 8–9 slides for a home + auto package.
• Recommendation on slide 2.
• Program value visible before detailed caveats.
• No standalone data-quality slide.
• Confidence shown once near the recommendation.
• No more than three decision blocks.
• No more than three coverage priorities.
• One primary CTA.
• Detailed source limitations in the final appendix.

Sub-Skill 2: Multi-Carrier Comparison

When to Use

Use only when two or more carrier/program options for the same insured are sufficiently comparable or can be compared with clearly disclosed caveats.

Required Inputs

• At least two documented options.
• Consistent insured, risk, household, vehicle, property, and effective-date assumptions.
• Carrier/program and quote status.
• Annualized premiums.
• Material limits, deductibles, forms, endorsements, fees, and underwriting conditions.

Advisor Decisions

Ask only:

1. Output mode, when unspecified.
2. Explicit recommendation or recommendation framework.
3. Recommended option only when the framework must

Truncated - read the full file at https://github.com/refocus-ai/refocus-ai-insurance-tools/blob/3cbf3aa319db9074dcaff9e53d86c871502029e6/skills/personal-lines-proposal-builder/SKILL.md.