How does the skill prevent AI hallucinations?

It performs systematic spot-checks, comparing generated claims against original source documents to ensure every fact is grounded in the provided evidence.

Does this replace human attorney review?

No, it is designed to flag 'Attorney Decision Needed' items and provide a clean draft, but human oversight remains essential for final legal judgment.

Can this skill fix errors automatically?

Yes, it generates corrected versions of sections for critical and major issues, replacing hallucinated facts with investigation markers or corrected data.

What kinds of files does it review?

It reviews Markdown output files and compares them against source materials like PDFs, DOCX, or consolidated text files within the workspace.

Legal QA Review & Anti-Hallucination

Name: Legal QA Review & Anti-Hallucination
Author: jdrodriguez

byjdrodriguez

•

セキュリティとテスト

Performs rigorous factual verification and citation integrity checks on legal analysis to prevent AI hallucinations.

This skill serves as a critical safety layer for AI-generated legal work, systematically cross-referencing output files against original source materials like case files and firm context. It identifies unsupported claims, validates legal citations, detects internal inconsistencies, and automatically flags potential hallucinations with severity ratings. By ensuring every factual assertion is grounded in evidence, it minimizes the risks of professional sanctions or malpractice associated with unverified AI outputs in legal proceedings.

主な機能

012 GitHub stars

02Severity-rated issue reporting categorized by Critical, Major, and Minor impact

03Evidence-based spot-check logging to track accuracy across output files

04Automated fix generation for factual inconsistencies and missing data

05Automated anti-hallucination factual verification against source materials

06Legal citation integrity checks with specific flagging for unverified authority

ユースケース

01Verifying that specific names, dates, and test results match raw evidence files

02Final validation of AI-generated legal briefs and case summaries before attorney review

03Ensuring internal consistency and chronological accuracy across multi-section legal reports

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jdrodriguez/legal-toolkit qa-review

For use in Claude.ai and ChatGPT

主な機能

012 GitHub stars

02Severity-rated issue reporting categorized by Critical, Major, and Minor impact

03Evidence-based spot-check logging to track accuracy across output files

04Automated fix generation for factual inconsistencies and missing data

05Automated anti-hallucination factual verification against source materials

06Legal citation integrity checks with specific flagging for unverified authority

ユースケース

01Verifying that specific names, dates, and test results match raw evidence files

02Final validation of AI-generated legal briefs and case summaries before attorney review

03Ensuring internal consistency and chronological accuracy across multi-section legal reports