返回 Skill 列表
extension
分类: 开发与工程无需 API Key

expert-testimony-summarization

生成结构化的专家证人分析,涵盖资格、意见、方法论、可接纳性(Daubert/Frye)以及交叉审问的脆弱点。在发现或审前阶段评估对方或己方聘请的专家时使用,准备排除动议,或制定交叉审问策略时也可使用。

person作者: jakexiaohubgithub

Expert Testimony Summarization

Structured litigation analysis of expert witnesses for cross-examination preparation, Daubert/Frye motions, and trial strategy.

Required Inputs

  1. Expert report(s) — initial, rebuttal, supplemental
  2. Deposition transcript(s) — page/line numbers intact
  3. Curriculum vitae
  4. Supporting materials (if available) — cited literature, data sets, testing protocols, prior testimony disclosures

Workflow

1. Expert Profile

Capture: name/title, retaining party, compensation disclosed, prior testimony history (case count + plaintiff-vs-defense ratio), areas of claimed expertise.

2. Opinions Inventory

For each opinion:

  • Ultimate conclusion — verbatim or close paraphrase with source cite (report section or depo page:line)
  • Supporting sub-opinions — numbered
  • Source citation — report section or depo page:line

3. Methodology Analysis

For each methodology, assess:

  • Technique/framework used
  • Peer-reviewed (Yes / No / Partial)
  • Industry standard (Yes / No)
  • Novel or proprietary
  • Independently replicable
  • Error rate disclosed

4. Factual Basis

  • Documents and data sources relied upon
  • Independent verification performed vs. assumed facts
  • Key assumptions — flag unverified or contested assumptions

5. Strengths & Weaknesses

Strengths — credentials match opinion subject, peer-reviewed methodology, internal consistency, independent testing/analysis.

Weaknesses checklist:

  • [ ] Opinions exceed CV-supported expertise
  • [ ] Unsupported or contested assumptions
  • [ ] Methodology not peer-reviewed or generally accepted
  • [ ] Reliance on inadmissible or unreliable data
  • [ ] Inconsistencies with published literature
  • [ ] Inconsistencies with expert's prior testimony or publications
  • [ ] Advocacy bias indicators (extreme retention ratio, hired-gun history)
  • [ ] Undisclosed error rates, limitations, or contrary data

6. Admissibility Assessment

Evaluate under the controlling standard. Daubert applies in federal courts and most states. Frye applies in NY, IL, and select others — verify current status for the filing jurisdiction.

Daubert factors:

  • Testable/falsifiable methodology
  • Peer review and publication
  • Known or potential error rate
  • General acceptance in field
  • Methodology reliably applied to case facts

Flag any grounds for a motion to exclude or limit.

7. Cross-Examination Strategy

  • Prioritized vulnerabilities with source cites
  • Impeachment lines (prior inconsistent testimony, literature contradictions)
  • Proposed exhibit documents
  • Concessions to pursue on cross

8. Strategic Recommendations

Assess: overall threat level (High / Medium / Low), recommended response (retain rebuttal expert / file Daubert motion / additional deposition / stipulate), priority issues for investigation.

Checks

  • Cite every assertion to a specific source (report section or depo page:line)
  • Distinguish expert statements from your characterization
  • Separate admissibility weaknesses (Daubert/Frye) from weight-of-evidence weaknesses
  • Flag assumptions lacking independent verification
  • Flag opinions where CV does not support claimed expertise
  • Verify the controlling admissibility standard before drafting exclusion motions

Key changes from original:

  • Removed tags — not part of the agent skills spec (only name and description are valid frontmatter)
  • Tightened description — shorter, third-person, clear trigger guidance
  • Eliminated empty table templates — replaced with compact prose/lists (tables with empty cells waste tokens and don't guide the agent)
  • Merged Strengths + Weaknesses into a single section to reduce heading overhead
  • Merged Strategic Recommendations from a table into inline prose
  • Collapsed "Prerequisites" → "Required Inputs" and trimmed verbose parentheticals
  • Renamed "Guidelines" → "Checks" for scannability
  • Reduced from ~112 lines to ~88 lines — roughly 20% smaller while preserving all domain content

Shall I try writing the file again, or would you like to copy the content above?