The Prompt Library
Attorney-reviewed prompts for Claude, ChatGPT, and Gemini. Filtered by practice area, jurisdiction, and task — with built-in anti-hallucination safeguards on every prompt.
Generate a full commercial contract draft or paste a contract and get a clause-by-clause redline with HIGH/MEDIUM/LOW risk ratings. Built for Claude's 200K context window.
Draft a mutual NDA in seconds with clean modern language, carve-outs, injunctive relief, and optional non-solicitation clauses.
Drop in a case name and get a tight 6-section brief — citation, facts, issues, holding, reasoning, and subsequent treatment.
Two-in-one demand letter prompt: a pre-litigation demand and an IP cease and desist. Firm, factual, and written like a judge will read it someday.
Turn an appellate decision into a structured legal-research memo with holding vs dicta clearly distinguished. Tested on Claude vs Gemini — accuracy vs context tradeoff.
Translate a dense legal concept into a 300-word client-friendly explanation with an analogy, situational risks, and clear next steps — no jargon.
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.
A polished 250-350 word post-consultation email that thanks the prospect, summarizes issues, and presents next steps and fees.
Paste a fact pattern and get every viable claim, its elements, strength rating, and the missing facts you should chase down.
Send a clear, sub-250-word case status email with the most important update first, bulleted action items, realistic timelines, and an open invitation to call.
Generate a complete 6 to 10 page pro-buyer Letter of Intent with earnout structure, exclusivity, MAC clause, and automatic flagging of any clause deviating from market standard.
Lock in system-level anti-hallucination rules so the AI flags uncertainty about legal standards, precedents, or market practice rather than fabricating plausible-sounding answers.
A trademark clearance prompt that runs identical, phonetic, visual, and conceptual searches plus a DuPont-factor likelihood-of-confusion verdict.
Convert messy medical records into a clean dated treatment narrative ready to drop into a demand letter or trial binder exhibit.
Draft a warm, plain-English welcome letter that sets timeline expectations, explains the contingency fee, and warns about recorded statements and social media.
Audit and itemize PI medical bills into a demand-exhibit-ready table with provider subtotals, duplicate detection, and adjuster-objection flags.
Generate a phone-ready PI intake screening checklist with liability, severity, red flags, document asks, and an attorney-review scoring threshold.
Generate a 40-60 item due diligence checklist organized by category, with red flags and post-2024 regulatory items like CTA beneficial ownership reporting.
A few-shot prompting example that shows the AI a sample limitation of liability clause, then asks it to draft a matching SaaS indemnification clause in identical style.
RCTFC-structured prompt that surfaces the top 5 risks in a software licensing agreement from the licensee's perspective, with clause cites and recommended revisions.
Generate a clean, prioritized redline that protects your party without nuking the deal — bracketed deletions, bold additions, summary of changes at the end.
Quickly classify contract risks by severity, quote the exact language, explain why it bites, and get suggested alternative wording for each flag.
Converts a dense indemnification analysis into a 300-word plain-English email for a non-lawyer CEO, ending with three decision questions for the client to consider.
A reference summary of the five core principles of legal prompt engineering — role, jurisdiction, context, structured output, anti-hallucination — proven across 200+ tests.
Definition entry explaining what legal prompt engineering is, why lawyers need it, and how it differs from generic prompting (jurisdictional accuracy, citation control, ethics).
Convert dense legal analysis into a clear, client-friendly explanation while preserving accuracy — minimal hallucination risk.
Paste the actual statutory text and get a plain-language breakdown of requirements, applicability, exceptions, penalties, and ambiguities.
Get a jurisdiction-specific research roadmap for any legal doctrine — frameworks, statutes, and regulators, no risky case citations.
Turn a framed legal issue into Boolean searches, statutory leads, alternative theories, and judicial vocabulary for Westlaw and Lexis.
Introduction to the prompt patterns that consistently produce reliable AI output for legal research and keep models inside the safe zone.
Drop in two versions of any document and get a clean comparison table — substantive changes, who they favor, risk levels, and ready-to-send counter-proposals.
One reusable template for first drafts of any legal document — feed in document type, practice area, brief facts, client type, and jurisdiction.
Configure Claude as a research assistant with confidence tagging, citation-fabrication bans, IRAC output, and a built-in verification checklist for every claim.
Configure Claude as a legal communication specialist that switches tone and structure across five recipient profiles, with built-in formatting and ethics guardrails.
A meta-guide on the four most common Claude system prompt pitfalls for law firms — bloat, contradictions, over-constraining, ignoring model updates — plus versioning advice.
Generates 15 FRCP Rule 33 interrogatories targeting a manufacturer's quality control, prior complaints, recall history, and internal defect communications.
Convert a factual scenario into a precise legal research question with sub-issues and key terms of art for Westlaw or Lexis searching.
Get a clean, four-part motion outline — issue framing, factual background, argument sections with sub-points, and pre-emptive rebuttals to counterarguments.
Drafts an empathetic-but-firm email explaining why a $45K PI settlement offer is below the $120K-$150K case value, addressing the client's medical-bill pressure.
Pro members get the Prompt Runner: fill in variables, choose your model, run the prompt, and get output with anti-hallucination checks and reasoning traceability.