Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Orchestrates a comprehensive multi-agent contract review that analyzes risk, plain-English translation, missing protections, and compliance in parallel. Use when a user shares a contract and wants a full review, safety score, or executive summary. Trigger with "/contract-review" or "review this contract".
.claude/skills/jeremylongshore-contract-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 109% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 5% | 0% |
Flagship contract review skill that spawns five parallel analysis agents, then aggregates their findings into a single executive report with a Contract Safety Score. Designed for founders, freelancers, and small-business operators who need to understand what they are signing without retaining outside counsel for every agreement.
Most contracts arrive as walls of dense legalese. A single-pass review misses nuance. This skill mirrors how a law firm reviews contracts: multiple specialists work in parallel, each with a different lens, then a senior partner synthesizes everything into a recommendation.
The orchestrator:
compliance, and party-balance analysis).
DOCX, TXT, MD) or pasted directly into the conversation.
Determine the contract type from one of the following categories:
Capture: parties, effective date, term/duration, governing law, dispute resolution mechanism, total contract value (if stated).
Each agent receives the full contract text and returns structured findings.
| Agent | Focus | Key Deliverable | |-------|-------|-----------------| | Risk Analyst | Clause-by-clause risk scoring across 10 categories | Risk heat map, poison pill flags | | Plain-English Translator | 8th-grade reading level rewrite | Clause-by-clause translation with flags | | Protection Auditor | Gap analysis against type-specific checklists | Missing protections with urgency ratings | | Compliance Checker | Regulatory alignment (GDPR, CCPA, labor law basics) | Compliance findings table | | Party-Balance Analyst | Fairness tilt between the parties | Asymmetry flags, one-sided clause list |
Combine all five agent reports into a unified document with these sections:
Weighted formula:
| Component | Weight | Source Agent | |-----------|--------|-------------| | Risk severity (inverse) | 30% | Risk Analyst | | Protection coverage | 25% | Protection Auditor | | Party balance | 20% | Party-Balance Analyst | | Compliance alignment | 15% | Compliance Checker | | Language clarity | 10% | Plain-English Translator |
Letter grades: A (90-100), B (80-89), C (70-79), D (60-69), F (< 60).
filename.
Filename: CONTRACT-REVIEW-{party-or-title}-{YYYY-MM-DD}.md
The report uses Markdown with tables and follows this structure:
# Contract Review Report
## Executive Summary
## Contract Metadata
## Contract Safety Score: [score]/100 ([grade])
## Risk Heat Map
## Plain-English Quick Reference
## Missing Protections
## Compliance Findings
## Party-Balance Assessment
## Recommended Next Steps
## Disclaimer| Failure Mode | Cause | Resolution | |--------------|-------|------------| | File not found | Path is incorrect or file missing | Ask the user to confirm the file path | | Unreadable format | Binary or encrypted document | Ask for a plain-text or PDF version | | Document too long | Exceeds context window | Summarize by section; warn about truncation | | Agent timeout | One parallel agent fails to return | Report partial results; note which agent failed | | Ambiguous contract type | Cannot classify confidently | Ask the user to confirm the contract type |
Example 1 — File-based review:
> User: Review the contract at ~/contracts/acme-msa-2026.pdf
The orchestrator reads the file, classifies it as a Master Services Agreement, launches five agents, and produces a report:
Contract Safety Score: 72/100 (C)
Top Risk: Unlimited indemnification liability (Section 8.2)
Missing: No force majeure clause, no data breach notification timeline
Balance: Tilts 65/35 in favor of Acme Corp
Recommendation: Negotiate Sections 8.2 and 12.1 before signingExample 2 — Pasted text:
> User: Review this contract: pasted NDA text]
The orchestrator classifies it as a Mutual NDA, flags a unilateral non-solicitation clause hidden in the definitions, and scores it 58/100 (F).
contract templates (CC BY 4.0) used as comparison baselines.
— Authoritative clause language references.
Institute's foundational contract law principles.
Privacy Act via the CA Attorney General.
Legal Disclaimer: This skill provides AI-generated analysis for informational and educational purposes only. It does not constitute legal advice, create an attorney-client relationship, or substitute for consultation with a qualified attorney. Contract interpretation depends on jurisdiction, context, and specific facts that an AI cannot fully evaluate. Always consult a licensed attorney before making legal decisions based on this analysis.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,690 | 9,053 | -38% | 1 | 1 | 0% | 2,600 | 2,456 | -6% | 0 | 0 | — |
case-02 | fail→fail | 33,863 | 6,589 | -81% | 1 | 1 | 0% | 6,222 | 2,223 | -64% | 0 | 0 | — |
case-03 | pass→pass | 12,372 | 8,469 | -32% | 1 | 1 | 0% | 2,106 | 3,127 | +48% | 0 | 0 | — |
case-08 | pass→pass | 4,617 | 1,845 | -60% | 1 | 1 | 0% | 920 | 1,973 | +114% | 0 | 0 | — |
case-04 | fail→pass | 7,472 | 1,848 | -75% | 1 | 1 | 0% | 1,168 | 1,881 | +61% | 0 | 0 | — |
case-05 | fail→pass | 7,779 | 1,884 | -76% | 1 | 1 | 0% | 1,210 | 1,973 | +63% | 0 | 0 | — |
case-06 | pass→pass | 6,086 | 2,002 | -67% | 1 | 1 | 0% | 1,065 | 2,004 | +88% | 0 | 0 | — |
case-07 | pass→pass | 3,724 | 1,756 | -53% | 1 | 1 | 0% | 655 | 1,951 | +198% | 0 | 0 | — |
case-09 | fail→pass | 5,089 | 2,203 | -57% | 1 | 1 | 0% | 996 | 2,079 | +109% | 0 | 0 | — |
case-10 | pass→pass | 8,540 | 1,685 | -80% | 1 | 1 | 0% | 1,424 | 1,880 | +32% | 0 | 0 | — |
case-11 | fail→pass | 7,521 | 1,603 | -79% | 1 | 1 | 0% | 1,340 | 1,938 | +45% | 0 | 0 | — |
case-12 | pass→pass | 7,121 | 2,089 | -71% | 1 | 1 | 0% | 1,208 | 2,031 | +68% | 0 | 0 | — |
case-13 | pass→pass | 9,028 | 5,370 | -41% | 1 | 1 | 0% | 1,544 | 2,498 | +62% | 0 | 0 | — |
case-14 | pass→pass | 5,912 | 2,168 | -63% | 1 | 1 | 0% | 873 | 1,957 | +124% | 0 | 0 | — |
case-15 | fail→pass | 11,886 | 4,201 | -65% | 1 | 1 | 0% | 2,237 | 2,357 | +5% | 0 | 0 | — |
case-16 | pass→pass | 10,383 | 6,390 | -38% | 1 | 1 | 0% | 1,743 | 2,864 | +64% | 0 | 0 | — |
case-17 | pass→pass | 12,191 | 7,611 | -38% | 1 | 1 | 0% | 1,998 | 2,875 | +44% | 0 | 0 | — |
case-18 | fail→pass | 12,331 | 2,872 | -77% | 1 | 1 | 0% | 2,267 | 2,089 | -8% | 0 | 0 | — |
case-19 | pass→pass | 8,224 | 2,438 | -70% | 1 | 1 | 0% | 1,439 | 1,956 | +36% | 0 | 0 | — |
case-20 | fail→fail | 10,259 | 1,436 | -86% | 1 | 1 | 0% | 1,655 | 1,820 | +10% | 0 | 0 | — |
case-21 | pass→pass | 10,068 | 3,146 | -69% | 1 | 1 | 0% | 1,662 | 2,193 | +32% | 0 | 0 | — |
case-22 | fail→fail | 28,168 | 22,684 | -19% | 1 | 1 | 0% | 5,201 | 5,838 | +12% | 0 | 0 | — |
case-23 | pass→pass | 6,831 | 4,316 | -37% | 1 | 1 | 0% | 1,112 | 2,351 | +111% | 0 | 0 | — |
case-24 | pass→pass | 7,348 | 4,845 | -34% | 1 | 1 | 0% | 1,270 | 2,424 | +91% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted, and 22 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +25 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.