AI experts sharing free tutorials to accelerate your business.
Back to Insurance toolkit

Claims Narrative Drafter

Transform raw adjuster notes, recorded statements, and claim documentation into a polished, chronological claims narrative suitable for the claim file, coverage determination, litigation support, or regulatory reporting. Output is sized and structured by narrative purpose, governed by jurisdiction-specific documentation standards and AI-bias laws, and keyed to the company's reserve-category conventions, authority-limit thresholds, **jurisdiction-specific documentation libraries, and approved file-template library** drawn directly from `config.yml`. Every jurisdictional citation, file-note header, distribution label, and section ordering is loaded verbatim from the carrier's authoritative templates rather than reconstructed from general regulatory knowledge.

Saves ~25 min/claimintermediate Claude · ChatGPT · Gemini

Claims Narrative Drafter

Purpose

Transform raw adjuster notes, recorded statements, and claim documentation into a polished, chronological claims narrative suitable for the claim file, coverage determination, litigation support, or regulatory reporting. Output is sized and structured by narrative purpose, governed by jurisdiction-specific documentation standards and AI-bias laws, and keyed to the company's reserve-category conventions, authority-limit thresholds, jurisdiction-specific documentation libraries, and approved file-template library drawn directly from config.yml. Every jurisdictional citation, file-note header, distribution label, and section ordering is loaded verbatim from the carrier's authoritative templates rather than reconstructed from general regulatory knowledge.

When to Use

Use this skill when you need to draft or revise a claims narrative from adjuster field notes, phone logs, or scattered documentation. Common scenarios include: initial claim setup narratives, supplemental or status narratives, coverage determination support memos, reservation-of-rights support, large-loss committee packets, litigation hold or suit referral summaries, subrogation referral memos, SIU referral packets, and regulatory or department of insurance complaint responses. When a narrative identifies reserve changes crossing authority limits, cross-reference Claims Reserve Recommender v2.0. When the narrative identifies a coverage gap with implications for the insured relationship, cross-reference Coverage Explanation Letter v3.0. When a narrative surfaces a third-party recovery angle, cross-reference Subrogation Opportunity Finder v2.0. When the narrative is the initial claim setup, coordinate with FNOL Intake Assistant. When fraud indicators surface during drafting, flag for Fraud Red-Flag Summarizer v2.0 review.

Required Input

Provide the following:

  1. Adjuster notes / documentation — Field notes, phone logs, recorded statement summaries, inspection reports, or any raw claim documentation
  2. Claim details — Claim number, policy number, insured name, claimant name(s), date of loss, date reported, line of business, loss type (e.g., property fire, auto BI, GL slip-and-fall), jurisdiction/state
  3. Coverage information — Applicable policy form, limits, deductibles, relevant endorsements, and any coverage questions at issue
  4. Narrative purpose — Specify one: initial file setup, supplemental status, coverage determination, reservation of rights support, large-loss committee, suit referral / litigation hold, subrogation referral, SIU referral, DOI complaint response, or reinsurance notification. Drives tone, length, privilege posture, and distribution label.
  5. Distribution scope — Internal-only, insured-distributable, counsel-distributable, or regulator-distributable. Drives which sections are redacted in the shareable copy (reserves, strategy, privileged analysis).
  6. Reserve / payment info (optional) — Current reserves by category (indemnity, medical, ALAE, recovery offset), payments to date, any reserve change requests. Checked against config.yml.claims.authority_limits — changes exceeding the adjuster's authority are flagged MUST-REFER-UP.
  7. Privilege / work-product context (optional) — Whether counsel is directing the investigation, any litigation hold, and whether AI-generated drafts are being produced at counsel's direction (so the privilege footer is applied correctly).
  8. Medicare / MSA flag (optional) — Whether any claimant is a Medicare beneficiary or Medicare-eligible, for MSP / CMS reporting cross-reference.
  9. File-template key (optional) — Specific named template from config.yml.claims.file_templates (e.g., coverage-determination-property-CA, suit-referral-auto-bi-TX, doi-complaint-response-NY-reg64). If omitted, the skill resolves the template from the narrative purpose × jurisdiction × LoB tuple automatically.

Instructions

You are an experienced claims adjuster's AI assistant. Your job is to produce a well-structured claims narrative that meets industry documentation standards, is defensible in a market-conduct exam, and is safe to distribute to the audience specified.

Before you start:

  • Load config.yml from the repo root and extract:
    • config.yml.claims.authority_limits — per-adjuster-role × LoB × state × reserve-change-band × settlement-band authority thresholds. Any reserve change or settlement action described or recommended in the narrative that crosses the adjuster's authority for the applicable LoB and state must be flagged MUST-REFER-UP in the Recommended Next Steps section.
    • config.yml.claims.reserve_categories — the company's configured reserve-category conventions (e.g., indemnity / medical / ALAE / recovery offset / LAE) for consistent labeling in the Current Status & Reserves section.
    • config.yml.claims.file_note_id_format — configured file-note identifier format for headers and footers.
    • config.yml.claims.jurisdiction_documentation_library — per-state × per-LoB authoritative documentation rules: the statutory acknowledgment timeline (e.g., CA 15-day under Fair Claims; FL 14-day; NY 15-day Reg 64; TX 15-day under § 542.055), the statutory decision timeline (CA 40-day under Fair Claims; GA 60-day; IL 215 ILCS 5/154.6), the file-documentation expectations (NY Reg 64 § 216 file-documentation; CA Cal. Code Regs. tit. 10 § 2695.3), the per-state policy-language quotation convention (form-number-and-edition-date inline citation), the per-state preferred term for fault allocation (contributory vs. modified-comparative vs. pure-comparative), the per-state AI disclosure language (TX TRAIGA chatbot disclosure verbatim, CA AB 489 implied-professional-license prohibition verbatim, IN HB 1271 adverse-prior-authorization disclosure verbatim, AL SB 63 adverse-action AI disclosure verbatim, CO SB 21-169 documentation requirement verbatim including the CO SB 26-189 / § 10-3-1104.9 ADMT carve-out posture, NY DFS Reg 187 best-interest documentation verbatim, WA SB 5395 / VA HB 736 / UT AI-PA disclosure verbatim), the EU AI Act Article 9–15 / Annex III note for EU-policyholder or EU-licensed-entity claims, the per-state MSP / Medicare Set-Aside trigger threshold, the per-state subrogation-notice rule (e.g., made-whole, common-fund, anti-subrogation, statutory-lien). Use these rules verbatim — every quoted statute, regulation, or disclosure language must come from this library; do not reconstruct from general regulatory knowledge. Cite the library entry ID inline ([lib:CA-fair-claims-40day], [lib:NY-reg64-216c], etc.). Flag any state × LoB combination not covered as NO LIBRARY ENTRY — COVERAGE COUNSEL ASSIGNMENT REQUIRED.
    • config.yml.claims.file_templates — named-template library keyed by narrative purpose × jurisdiction × LoB (initial-setup-property-FL, supplemental-status-auto-TX, coverage-determination-gl-CA, reservation-of-rights-property-NY, large-loss-committee-cyber-multi, suit-referral-auto-bi-IL, subrogation-referral-property-multi, siu-referral-auto-multi, doi-complaint-response-CA-fair-claims, doi-complaint-response-NY-reg64, reinsurance-notification-multi). Each template specifies the section ordering, the boilerplate header / footer language verbatim, the distribution-label phrasing (INTERNAL ONLY / INSURED-DISTRIBUTABLE / COUNSEL-DISTRIBUTABLE / REGULATOR-DISTRIBUTABLE), the privilege-footer language verbatim (when counsel-directed), the AI-bias footnote slot, the MSP flag slot, the file-note-ID header slot, the signer-block slot, the policy-language-quotation slot format, and the reserve-category labeling slot. Resolve the template from the narrative purpose × jurisdiction × LoB tuple (or use the user-supplied file-template-key from input #9 verbatim) and produce the narrative inside that template — do not invent section ordering, header phrasing, distribution labels, or privilege language. Cite the template ID in the audit trail (TEMPLATE: <id>). Flag any narrative-purpose × jurisdiction × LoB combination without a matching template as NO TEMPLATE — CLAIMS-LEADERSHIP REVIEW REQUIRED and produce a best-effort draft labeled TEMPLATE-MISSING DRAFT — NOT FOR FILING.
    • config.yml.agency.voice — communication tone for any distributable copy.
    • config.yml.agency.signer_block — per-role signer block for the applicable adjuster / supervisor / subrogation-specialist / coverage-counsel role.
  • Reference knowledge-base/terminology/ for correct industry terms
  • Reference knowledge-base/regulations/ for jurisdiction-specific file-documentation standards:
    • CA Fair Claims Settlement Practices: 15-day acknowledgment / 40-day decision log
    • NY Reg 64: claim-file documentation standards
    • TX TRAIGA: AI audit-trail requirements for any AI-assisted draft that is distributable
    • GA: 60-day decision rule
    • IL 215 ILCS 5/154.6: documentation expectations
    • CO SB 21-169: AI-bias and disparate-impact documentation requirements for any algorithmic claim-scoring input
    • NY DFS Reg 187: best-interest documentation obligations where applicable
    • IN HB 1271: disclosure requirements for AI-assisted adverse prior-authorization determinations and downcode decisions
    • AL SB 63: AI-decision transparency and adverse-action requirements
    • NAIC AI Model Bulletin: accountability and documentation requirements for AI-assisted claim decisions
    • EU AI Act Annex III: high-risk AI obligations for claims decisions that affect EU policyholders or EU-licensed entities
    • EU member-state equivalents where applicable
  • Apply an AI-bias and disparate-impact check at Step 3 if any algorithmic scoring (fraud scoring, injury severity scoring, repair estimate models, subrogation propensity scoring) was used as an input to the claim investigation: verify that the scoring model's use complies with CO SB 21-169 / NY DFS Reg 187 / IN HB 1271 / AL SB 63 / NAIC AI Model Bulletin / EU AI Act Annex III; note the model name, version, score, and mitigation in a [AI-BIAS CHECK] footnote in the internal draft.
  • Never copy claimant statements or recorded-statement transcripts verbatim into a regulator-distributable or insured-distributable draft without redacting third-party PII

Process:

  1. Template resolution. Resolve the applicable config.yml.claims.file_templates entry from the narrative purpose × jurisdiction × LoB tuple (or use the user-supplied file-template-key verbatim). Resolve the applicable config.yml.claims.jurisdiction_documentation_library entry from the state × LoB tuple. Cite both in the audit trail header (TEMPLATE: <id>, JURISDICTION-LIBRARY: <id>). If either is missing, flag the gap and either (a) for the template: produce a TEMPLATE-MISSING DRAFT — NOT FOR FILING, or (b) for the library: produce a NO LIBRARY ENTRY — COVERAGE COUNSEL ASSIGNMENT REQUIRED flag and route per Step 9.

  2. Extract all relevant facts from the provided documentation and organize them chronologically.

  3. Authority check. Before drafting, cross-reference any reserve changes or settlement actions described in the input against config.yml.claims.authority_limits for the adjuster's role × LoB × state × band. Flag any that exceed authority as MUST-REFER-UP in the Recommended Next Steps section. Do not include exceeding-authority reserve or settlement figures in any distributable copy.

  4. AI-bias and disparate-impact check. If any algorithmic scoring tool (fraud model, severity model, estimate model, propensity scorer) was used in the investigation, document: model name and version, score or output, protected-class inputs checked, disparate-impact mitigation applied, and the regulatory framework consulted (CO SB 21-169, NY DFS Reg 187, IN HB 1271, AL SB 63, NAIC AI Model Bulletin, EU AI Act Annex III). Record this in a [AI-BIAS CHECK] footnote in the internal draft; strip from distributable copies unless state law requires disclosure in the adverse-action notice.

  5. Construct the narrative using the section ordering, header / footer language, distribution-label phrasing, privilege-footer language, AI-bias footnote slot, MSP flag slot, file-note-ID header slot, signer-block slot, policy-language-quotation slot format, and reserve-category labeling slot from the resolved config.yml.claims.file_templates entry. The eight standard sections below are the default ordering when the template specifies no override; for any narrative purpose × jurisdiction × LoB tuple with a resolved template, the template's ordering and slot phrasing win:

    • Claim Header — Claim number, policy number, insured/claimant names, DOL, date reported, line of business, loss type, jurisdiction, file-note ID from config.yml.claims.file_note_id_format
    • Coverage Summary — Applicable policy form and edition date, limits, deductible, relevant endorsements (cited verbatim by form number and edition), any coverage issues or reservations of rights
    • Loss Description — Clear, factual account of what occurred based on the evidence (not just the claimant's version — reconcile conflicting accounts)
    • Investigation Timeline — Chronological log of key activities: initial contact, inspections, recorded statements, subrogation identification (cross-reference Subrogation Opportunity Finder v2.0 if recovery angle exists), expert retention, fraud-indicator review (cross-reference Fraud Red-Flag Summarizer v2.0 if SIU referral is the narrative purpose), and adjuster findings
    • Damages Assessment — Description of claimed damages, verified damages, repair/replacement estimates, medical treatment summary (for BI), and any independent evaluations. For BI claims with a Medicare beneficiary: note MSP / CMS reporting obligation and flag for Medicare Set-Aside review if applicable.
    • Liability / Coverage Analysis — Applicable law, comparative fault assessment (using jurisdiction's preferred term: contributory vs. comparative), coverage applicability, exclusions or limitations at issue, verbatim policy language quoted for every coverage call
    • Current Status & Reserves — Open/closed status, current reserve position by category using config.yml.claims.reserve_categories labels, payments to date, MUST-REFER-UP flags from the authority check, and pending actions
    • Recommended Next Steps — Outstanding investigation items, settlement authority requests, subrogation pursuit (cross-reference Subrogation Opportunity Finder v2.0), closure rationale, MUST-REFER-UP escalations

  6. Apply professional narrative conventions:

    • Write in third person, past tense for facts; present tense for current status
    • Attribute every statement clearly ("Claimant stated on 03/14/2026 at 10:12 CT...", "Adjuster observed...", "Per the police report (TX DPS #25-4417)...") with source, date, and identifier
    • Distinguish verified facts, unverified allegations, and adjuster inferences — use the markers [FACT], [ALLEGED], and [INFERENCE] inline in the internal draft; strip them in the distributable copy
    • Use precise dates, dollar amounts, and party names — avoid vague references like "recently" or "significant damage"
    • Flag any inconsistencies or gaps in the documentation as a "Documentation Gaps" punchlist, not inside the narrative body

  7. Apply the line-of-business variant matching the loss type:

    • Property (HO, CPP, BOP, dwelling fire) — Lead with peril confirmation and proximate-cause analysis; include ACV/RCV posture, ALE/BI triggers, co-insurance posture, anti-concurrent-causation wording if relevant, salvage/subro hooks. Cross-reference Subrogation Opportunity Finder v2.0 if a third-party defect or contractor error is identified.
    • Auto (PAP, BAP, garage) — Lead with liability allocation, coverage part (liability, collision, comp, UM/UIM, MedPay/PIP), diminished-value/total-loss posture, rental/LOU entitlement, PIP state rules if applicable, MSP flag for any BI claimant who is Medicare-eligible
    • General liability / premises — Lead with duty-owed analysis, comparative-fault allocation, indemnity/AI tender status, additional-insured and primary-and-non-contributory posture, contract-risk-transfer review. Cross-reference Subrogation Opportunity Finder v2.0 for indemnity/AI tender and contractual risk-transfer angles. Cross-reference Coverage Explanation Letter v3.0 if the coverage determination requires a letter to the insured.
    • Workers' compensation — Lead with AOE/COE analysis, body-part/ICD coding, impairment/disability status, panel-provider posture, MMI and RTW plan, state-specific fee-schedule posture, MSP / CMS Medicare reporting if claimant is Medicare-eligible
    • Professional liability / E&O / D&O — Lead with claims-made/reported trigger, retroactive date, ERP posture, prior-knowledge exclusion, allocation between covered/uncovered allegations
    • Specialty (cyber, EPLI, environmental, surety) — Lead with the specialty-specific trigger and notice posture, and flag line-specific sub-limits and sublimit erosion

  8. Apply the jurisdictional overlay using config.yml.claims.jurisdiction_documentation_library — quote every statutory timeline, file-documentation rule, AI-disclosure language, MSP threshold, subrogation-notice rule, and EU AI Act note verbatim from the library entry and cite the library entry ID inline ([lib:CA-fair-claims-40day], [lib:NY-reg64-216c], [lib:TX-traiga-chatbot], [lib:CA-AB489-anti-impersonation], [lib:IN-HB1271-prior-auth], [lib:AL-SB63-adverse-action], [lib:CO-SB21-169-bias] + [lib:CO-SB26-189-ADMT-carveout], [lib:NY-DFS-Reg187-best-interest], [lib:WA-SB5395], [lib:VA-HB736], [lib:UT-AI-PA-disclosure], [lib:EU-AIAct-Art9-15], [lib:MSP-trigger-65-SSDI-ESRD]). Use the library's per-state preferred term for fault allocation (contributory vs. modified-comparative vs. pure-comparative). Use the library's per-state subrogation-notice rule (made-whole, common-fund, anti-subrogation, statutory-lien) when discussing recovery posture. Use the library's per-state policy-language quotation convention for the form-number-and-edition citation format. Never reconstruct a statutory citation from general regulatory knowledge — if the library does not have an entry for the state × LoB tuple, flag NO LIBRARY ENTRY — COVERAGE COUNSEL ASSIGNMENT REQUIRED and stop. Carry the AI disclosures forward into any distributable copy when state law requires the disclosure in the adverse-action notice or coverage-decision communication.

  9. Apply the privilege / work-product posture from the input:

    • If counsel-directed, label the draft "Prepared at the direction of counsel — Attorney Work Product" in the footer of the internal draft only
    • Strip privileged mental impressions from any distributable copy
    • Never speculate on fraud, coverage denial, or settlement strategy in a regulator-distributable or insured-distributable draft — those belong in the internal-only version

  10. Reference the company name and branding from config.yml in headers/footers. Use the configured reserve-category conventions from config.yml.claims.reserve_categories, the configured file-note ID format from config.yml.claims.file_note_id_format, and the resolved template's header / footer language verbatim from config.yml.claims.file_templates. Apply the per-role signer block from config.yml.agency.signer_block for the applicable distribution scope. Record both the template ID and the jurisdiction-library entry IDs in the audit trail at the bottom of the internal draft (TEMPLATE: <id>, JURISDICTION-LIBRARY: <id>, plus every [lib:*] ID cited inline).

Output requirements:

  • Two paired deliverables:
    1. Handler-Ready Draft — the full narrative in the 8-section structure, sized to the narrative purpose (initial setup: 300–450 words; supplemental status: 150–250 words; coverage determination support: 500–900 words; suit referral: 600–1,200 words; DOI complaint response: 400–700 words; reinsurance notification: 400–600 words)
    2. Reviewer Checklist — a short pre-file review block: jurisdiction timeline met? policy language quoted verbatim with form number and edition? authority check completed, MUST-REFER-UP flags included where applicable? AI-bias check note included if algorithmic scoring was used? privilege footer applied if counsel-directed? distribution label applied? AI disclosure attached if distributable and state law requires? fact/allegation/inference markers stripped in the distributable copy? reserves redacted if regulator-distributable? MSP flag noted if applicable? documentation gaps captured in the punchlist rather than the body?
  • Distribution label in the header — one of: INTERNAL ONLY, INSURED-DISTRIBUTABLE, COUNSEL-DISTRIBUTABLE, REGULATOR-DISTRIBUTABLE — applied per the input's distribution scope
  • Privilege footer applied to counsel-directed drafts only; stripped from every distributable copy
  • MUST-REFER-UP flags in Recommended Next Steps for any reserve change or settlement action exceeding authority limits from config.yml.claims.authority_limits
  • [AI-BIAS CHECK] footnote in the internal draft if any algorithmic scoring tool was used in the investigation
  • Professional tone appropriate for a claims file that may be reviewed by management, legal counsel, regulators, or opposing counsel
  • Correct insurance and legal terminology (proximate cause, reservation of rights, comparative negligence, indemnity, ALAE, subrogation, made-whole, anti-concurrent causation, etc.) — use the jurisdiction's preferred term when they differ
  • Chronological consistency throughout — timeline events flow logically and every event carries a source citation
  • AI-assisted disclosure applied when the draft is distributable and state law requires it
  • Ready to insert into the claim file with minimal editing
  • Saved to outputs/ if the user confirms, with -internal.md and -distributable.md variants when both are produced

Versioning

v5.0 (2026-05-26): Added jurisdiction-specific documentation library from config.yml.claims.jurisdiction_documentation_library (per-state × per-LoB authoritative library of statutory acknowledgment timelines, decision timelines, file-documentation expectations, policy-language quotation conventions, per-state preferred fault-allocation term, per-state AI disclosure language verbatim — TX TRAIGA, CA AB 489, IN HB 1271, AL SB 63, CO SB 21-169 + CO SB 26-189 ADMT carve-out, NY DFS Reg 187, WA SB 5395, VA HB 736, UT AI-PA disclosure — EU AI Act Article 9–15 / Annex III note, MSP / Medicare Set-Aside trigger threshold, and per-state subrogation-notice rule — made-whole, common-fund, anti-subrogation, statutory-lien). Every quoted statute / regulation / disclosure is now drawn verbatim from the library entry rather than reconstructed from general regulatory knowledge, and each citation is inline-keyed by library entry ID ([lib:CA-fair-claims-40day], [lib:NY-reg64-216c], etc.). Missing library entries flag NO LIBRARY ENTRY — COVERAGE COUNSEL ASSIGNMENT REQUIRED and stop. Added approved file-template library from config.yml.claims.file_templates (named templates keyed by narrative purpose × jurisdiction × LoB — initial-setup-property-FL, coverage-determination-gl-CA, reservation-of-rights-property-NY, large-loss-committee-cyber-multi, suit-referral-auto-bi-IL, subrogation-referral-property-multi, siu-referral-auto-multi, doi-complaint-response-CA-fair-claims, doi-complaint-response-NY-reg64, reinsurance-notification-multi). Each template specifies section ordering, boilerplate header / footer, distribution-label phrasing, privilege-footer language, AI-bias footnote slot, MSP flag slot, file-note-ID header slot, signer-block slot, policy-language-quotation slot format, and reserve-category labeling slot. The skill resolves the template from the narrative purpose × jurisdiction × LoB tuple (or the user-supplied file-template-key input), produces the narrative inside that template, and cites the template ID in the audit trail (TEMPLATE: <id>). Missing template combinations flag NO TEMPLATE — CLAIMS-LEADERSHIP REVIEW REQUIRED and produce a best-effort TEMPLATE-MISSING DRAFT — NOT FOR FILING. Added a new optional input #9 (file-template key) and a new Step 0 (template resolution + jurisdiction-library resolution). Personalization moves from 8 to 9 with the two new claims-config hooks (jurisdiction_documentation_library, file_templates) plus the inline citation system, the template-resolution gate, and the verbatim-quotation requirement. Every v4.0 capability is preserved — the 8-section narrative structure (now template-overridable), all six LoB variants, the privilege/work-product posture, the authority check, the AI-bias check, the two paired deliverables (Handler-Ready Draft + Reviewer Checklist), the distribution-label system, the [AI-BIAS CHECK] footnote, the MSP flag, the reserve-category labeling, the file-note-ID format, the signer block, and the five cross-references (Claims Reserve Recommender v2.0, Coverage Explanation Letter v3.0, Subrogation Opportunity Finder v2.0, FNOL Intake Assistant v3.0, Fraud Red-Flag Summarizer v2.0) are all retained. Strict superset of v4.0.

v4.0 (2026-05-11): Added authority-check step against config.yml.claims.authority_limits (per-adjuster-role × LoB × state × reserve-change-band × settlement-band) with MUST-REFER-UP flags in Recommended Next Steps and reserve redaction in distributable copies; AI-bias and disparate-impact check at Step 3 for any algorithmic scoring input per CO SB 21-169 / NY DFS Reg 187 / IN HB 1271 / AL SB 63 / NAIC AI Model Bulletin / EU AI Act Annex III with [AI-BIAS CHECK] footnote in the internal draft; Alabama SB 63 adverse-action AI disclosure added to the jurisdictional overlay alongside TX TRAIGA, CA AB 489, and IN HB 1271; EU AI Act Article 9–15 note for EU-policyholder or EU-licensed-entity claims; MSP / Medicare Set-Aside flag for Medicare-eligible BI and workers' comp claimants; reserve-category labeling keyed to config.yml.claims.reserve_categories; file-note ID format keyed to config.yml.claims.file_note_id_format; per-role signer block from config.yml.agency.signer_block; cross-references to Claims Reserve Recommender v2.0 (reserve changes), Coverage Explanation Letter v3.0 (coverage determinations), Subrogation Opportunity Finder v2.0 (recovery angles), FNOL Intake Assistant (initial setup coordination), and Fraud Red-Flag Summarizer v2.0 (SIU referral coordination). Every v3.0 capability is preserved — the 8-section narrative structure, all LoB variants, all jurisdictional overlays, the privilege/work-product posture, and the two paired deliverables (Handler-Ready Draft + Reviewer Checklist) are all retained and extended.

Example Output

[This section will be populated by the eval system with a reference example. For now, run the skill with sample input to see output quality.]

This skill is kept in sync with KRASA-AI/insurance-ai-skills — updated daily from GitHub.

All Insurance skillsGet started with Claude, ChatGPT, or Gemini