Agent Skillssamarailly51-pixel/claimpilot-harness › bodily-injury-claims-processing

bodily-injury-claims-processing

GitHub

用于审核和结构化汽车保险人身伤害索赔。涵盖证据完整性、伤情因果、治疗时序及赔偿支持分析,生成审查结论并指导下一步安全操作(批准/调查/拒绝),同时包含严格的合规与安全护栏。

skills/bodily-injury-claims-processing/SKILL.md samarailly51-pixel/claimpilot-harness

Trigger Scenarios

分析人身伤害索赔文件 设计理赔代理工作流 起草审查清单 识别缺失文档 推荐安全的索赔后续行动

Install

npx skills add samarailly51-pixel/claimpilot-harness --skill bodily-injury-claims-processing -g -y
More Options

Use without installing

npx skills use samarailly51-pixel/claimpilot-harness@bodily-injury-claims-processing

指定 Agent (Claude Code)

npx skills add samarailly51-pixel/claimpilot-harness --skill bodily-injury-claims-processing -a claude-code -g -y

安装 repo 全部 skill

npx skills add samarailly51-pixel/claimpilot-harness --all -g -y

预览 repo 内 skill

npx skills add samarailly51-pixel/claimpilot-harness --list

SKILL.md

Frontmatter
{
    "name": "bodily-injury-claims-processing",
    "description": "Review and structure auto insurance bodily injury claims, including intake triage, evidence completeness, injury causation, treatment chronology, medical necessity, wage-loss support, negotiation risks, and human escalation. Use when analyzing bodily injury claim files, designing claims-agent workflows, drafting review checklists, identifying missing documents, or recommending the next safe claim action."
}

Bodily Injury Claims Processing

Use a safety-first workflow. Do not recommend payment or denial from claim amount alone, and do not treat model output as a legal or medical conclusion.

Review workflow

  1. Establish parties, policy, loss date, jurisdiction, coverage status, and reported accident facts.
  2. Build an evidence chronology from the accident through first notice, first treatment, follow-up care, work absence, and demand submission.
  3. Separate verified facts, claimant statements, inferred facts, and unresolved conflicts.
  4. Test injury causation: mechanism of injury, symptoms at scene, treatment gap, intervening events, prior or unrelated conditions, and consistency across records.
  5. Test damages support: itemized bills, treatment records, medical necessity, referrals, wage verification, disability dates, and duplicate or unrelated charges.
  6. Classify the next action as approve, investigate, or deny. Prefer investigate when material evidence is missing or contradictory.
  7. Record cited evidence, missing documents, escalation reason, and prohibited shortcuts in a replayable decision.

Guardrails

  • Never auto-approve solely because a claim is below a monetary threshold.
  • Never infer causation from diagnosis or billing alone.
  • Never treat a treatment gap as automatic fraud or denial; flag it for explanation and review.
  • Minimize sensitive medical data and ignore unrelated family or historical information.
  • Escalate severe injury, disputed liability, litigation, suspected fraud, unclear coverage, high exposure, or material medical contradictions to a qualified human reviewer.
  • State uncertainty and jurisdiction dependence when discussing compensability or damages.

Output contract

Return a structured review containing:

  • recommended action and confidence
  • material findings
  • missing documents
  • cited evidence IDs
  • chronology conflicts
  • escalation reason
  • privacy or safety flags

Read references/review-checklist.md when performing a full file review or creating an evaluation case.

Version History

  • 8ec38df Current 2026-07-19 18:25

Metadata

Files
0
Version
d83a89b
Hash
4b8cf534
Indexed
2026-07-19 18:25

trang chủ - Wiki
Copyright © 2011-2026 iteam. Current version is 2.155.2. UTC+08:00, 2026-08-05 16:48
浙ICP备14020137号-1 $bản đồ khách truy cập$