Claude Prompt: Clause-by-Clause Contract Review with Reasoning Chains
Walk through every clause with explicit reasoning, a 1-10 risk score, suggested redlines, and a confidence rating you can verify line by line.
The Prompt
Variables (1)
| Variable | Description | Example |
|---|---|---|
| {{governing_law_of_the_contract}} | Governing Law of the Contract | State of Delaware |
Expected Output
A clause-by-clause breakdown with risk scores (1-10), four-step reasoning chains, recommendations, suggested redline language, and HIGH/MEDIUM/LOW confidence ratings, ending with a final summary of high-risk clauses and missing protections.
Usage Notes
Run this after extracting a structured intake brief. Force Claude to flag uncertain citations with [VERIFY: ...] tags so a supervising attorney has a clear audit trail to validate before the redline goes to the client.
Legal Sources Referenced
- ABA Formal Opinion 512 (2024) — Generative AI Tools and Lawyer Duties
- ABA Model Rule 1.1 (Competence) cmt. 8 — Technology competence
- Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) — sanctions for AI hallucinated citations
Originally featured in: How to Build a Claude Legal Workflow: From Client Intake to Final Document
Related Prompts
AI Prompt: Contract Risk Analysis Dashboard with Heat Map
Aggregate a clause review into a CEO-ready dashboard — heat map, top three risks with financial ranges, negotiation priorities, and a three-sentence executive summary.
Claude Prompt: Comprehensive Research Memo + Statute Analysis
Two heavy-research prompts in one: an Am Law 100 research memorandum and a section-by-section statute and regulation breakdown with compliance obligations.
Claude Prompt: Client Advisory Letter + Case Status Update Email
Two client-comm prompts in one: a sophisticated 1-2 page advisory letter and a sub-500 word case status email with bulleted action items.
Claude Prompt: Legal Research Assistant with Strict Anti-Hallucination Rules
Configure Claude as a research assistant with confidence tagging, citation-fabrication bans, IRAC output, and a built-in verification checklist for every claim.