Review Contracts for Legal Risk and Redlines1 skill
Analyzes legal contracts to surface risk, extract terms, suggest redlines, and compare provisions to market standards.
Legal Blueprint
A single skill that analyzes legal contracts, identifies unfavorable terms, suggests redlines, and compares provisions to market standards.
Run this command to deploy the blueprint to your environment.
Analyzes legal contracts and outputs risk assessments with severity ratings, red flags, key terms, market standard benchmarks, negotiability ratings, specific redline language, and missing provisions. It gives position-aware review (adjusting what counts as risky based on which party you represent) and document-type checklists grounded in the CUAD dataset of 41 legal risk categories, eliminating generic one-size-fits-all analysis that treats an NDA the same as an M&A agreement.
Once installed in your AI coding assistant (Claude Code, Cursor, GitHub Copilot, or 26+ compatible tools), you paste or reference a contract and specify your position (customer, vendor, buyer, seller, etc.). The skill analyzes the document for red flags, extracts key terms, compares provisions to market standards, flags internal inconsistencies, and outputs structured redlines in JSON. You review the output, then use the paired legal-redline-tools package to generate tracked-changes Word docs and redline PDFs. Output: a position-aware contract review with negotiation priorities and specific redline language ready to send.
Analyzes legal contracts to surface risk, extract terms, suggest redlines, and compare provisions to market standards.
First-pass contract reviews surface real negotiation priorities (not generic checklists) because analysis adjusts for your position and contract type, so you know exactly what to push back on.
Compatible with Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, and 26+ Agent Skills-compatible tools. Clone to your tool's skills directory. Built on CUAD dataset (41 legal risk categories from 510 real contracts), ContractEval benchmarks, and LegalBench. Pairs with legal-redline-tools package (pip install) to generate tracked-changes Word docs and PDFs from skill output.
Not legal advice. Qualified counsel should review material terms on any deal. Accuracy is approximately 0.62 F1 score (ContractEval benchmark) for clause extraction; best for first-pass review and issue flagging, not a replacement for attorney review on high-stakes transactions. Very long contracts may require section-by-section analysis due to context window limits. US law focus; enforceability varies by jurisdiction.
Position-Aware Review
Adjusts risk flagging based on your party role (customer, vendor, buyer, seller, licensor, licensee, receiving party). Same term has different risk profiles depending on which side you're on.
Red Flags Quick Scan
Instant detection of danger signs (liability cap <6 months, uncapped indemnification, unilateral amendment rights, perpetual obligations, offshore jurisdiction). Scans 12+ critical patterns.
Document Type Checklists
Specialized checklists for NDA, SaaS/MSA, Payment/Merchant, M&A, and Finder/Broker agreements. Checks contract-specific terms instead of applying one-size-fits-all criteria.
Market Standard Benchmarks
Compares terms (liability cap, non-compete, auto-renewal notice, escrow duration, SLA uptime) to industry standards with yellow/red thresholds. Shows you what's negotiable vs. what's market-rate.
Specific Redline Language
Outputs actual replacement text and suggested language for missing provisions, not just 'negotiate this.' Redlines export as structured JSON for Word/PDF generation.