Review contracts against your organization's negotiation playbook — flag deviations, generate confidence-scored redlines, and produce an auditable analysis with business impact assessment. Use when reviewing vendor contracts, customer agreements, SaaS agreements, professional services contracts, licenses, partnerships, procurement agreements, or any commercial agreement requiring clause-by-clause analysis against standard positions. Jurisdiction-agnostic with localization markers for any governing law. Supports playbook-based or general-standard review.
How this skill is triggered — by the user, by Claude, or both
Slash command
/legalcode-claude-code:legalcode-contract-reviewclaude-opus-4-5-20251101general-purposeThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use Legalcode MCP for source lookup while keeping documents and matter context in the user's agent environment.
Use Legalcode MCP for source lookup while keeping documents and matter context in the user's agent environment.
https://mcp.legalcode.md for anonymous laws and case law.https://mcppro.legalcode.md for stronger search, up to 20 results per query, guidance, agreements, downloads, and authenticated higher-throughput access.Disclaimer: This skill provides a framework for AI-assisted contract review. It does not constitute legal advice. All outputs should be reviewed by a qualified legal professional licensed in the relevant jurisdiction before use. Laws change; verify current applicability before relying on any provision described here. Statutory and case law references cited from memory carry hallucination risk — verify against authoritative sources before relying on them.
This skill reviews existing contracts against an organization's negotiation playbook or general commercial standards. It identifies deviations, classifies their severity, generates actionable redline suggestions, and produces a confidence-scored, auditable analysis.
Covers:
Does not:
This is a jurisdiction-agnostic skill. The contract's own governing law clause determines which legal framework applies. The review identifies the governing law early and adapts the analysis accordingly.
[JURISDICTION-SPECIFIC] When localizing, research and apply:
This skill uses interactive clarification at key decision points. Rather than assuming the user's intent, the workflow pauses and asks when:
Use the ask-user-question pattern (structured options with descriptions) wherever marked with ⟁ CLARIFY below. If the user has already provided the information (e.g., in the initial prompt or prior conversation), skip the question and proceed.
Accept the contract in any of these formats:
If no contract is provided, prompt the user to supply one.
⟁ CLARIFY — Before beginning the review, ask the user these questions. Present them as a structured set of options where possible:
Which side are you on?
Deadline: When does this need to be finalized?
Focus areas: Any specific concerns?
Deal context: Relevant business context?
If the user provides partial context, proceed with what you have but state your assumptions explicitly (e.g., "I'm assuming you are on the customer side — let me know if that's wrong and I'll re-run the analysis"). Do not silently assume.
Check for the organization's contract review playbook in local settings (e.g., legal.local.md or similar configuration files).
The playbook should define:
If no playbook is configured:
⟁ CLARIFY — Inform the user that no playbook was found, and ask which approach to take:
If proceeding without a playbook, clearly label the review: "Based on general commercial standards — not organizational positions"
Read the contract's governing law clause and identify the applicable jurisdiction. Then use legalcode-mcp to build a working legal reference file for this review.
legalcode-mcp is the Legalcode law, regulatory, and case law database with powerful search. Use it to gather the legal authority you need before analyzing clauses.
Research process:
Identify the governing law from the contract (e.g., "laws of the State of Delaware," "English law," "Polish law under the Kodeks cywilny").
⟁ CLARIFY — If any of the following are true, ask the user before proceeding:
Search legalcode-mcp for jurisdiction-relevant authority across these categories:
Save the most relevant results into a local temporary reference file (e.g.,
/tmp/legalcode-review-authority.md). Structure it as:
# Legal Authority Reference — [Contract Name]
## Governing Law: [Jurisdiction]
## Date Gathered: [date]
### Statutes & Regulations
- [Statute name, section, current text or summary]
- [...]
### Case Law
- [Case name, citation, key holding relevant to this contract]
- [...]
### Regulatory Guidance
- [Regulator, guidance title, relevance]
- [...]
Use this reference file throughout the analysis. When citing a statute or case in your clause analysis, pull from the verified reference file rather than from memory. Mark any citation sourced from legalcode-mcp as VERIFIED in the Glass Box audit trail.
If legalcode-mcp is not connected:
external_tools_used: "None — manual verification required"[JURISDICTION-SPECIFIC] For each jurisdiction, the research should cover at minimum:
Analyze the contract systematically. Read the entire contract before flagging issues — clauses interact with each other (e.g., an uncapped indemnity may be partially mitigated by a broad limitation of liability).
⟁ CLARIFY — For long or complex contracts (50+ pages, or with multiple schedules, exhibits, or incorporated documents):
Cover all 14 clause categories below (unless the user opted for a priority review). For each, assess against the playbook (or general standards) and note whether the clause is present, absent, or unusual.
| # | Clause Category | Depth | Key Review Points |
|---|---|---|---|
| 1 | Limitation of Liability | Deep | Cap amount, carveouts, mutual vs. unilateral, consequential damages |
| 2 | Indemnification | Deep | Scope, mutual vs. unilateral, cap, IP infringement, data breach, procedure |
| 3 | Intellectual Property | Deep | Pre-existing IP, developed IP, assignment, license grants, moral rights |
| 4 | Data Protection | Deep | DPA requirement, processing terms, sub-processors, breach notification, transfers |
| 5 | Term & Termination | Deep | Duration, renewal, convenience termination, cause termination, wind-down |
| 6 | Governing Law & Disputes | Deep | Jurisdiction, venue, arbitration vs. litigation, escalation |
| 7 | Confidentiality | Standard | Scope, term, carveouts, return/destruction obligations |
| 8 | Representations & Warranties | Standard | Scope, disclaimers, survival period, remedies for breach |
| 9 | Insurance | Standard | Coverage requirements, minimums, evidence of coverage |
| 10 | Assignment | Standard | Consent requirements, change of control, exceptions |
| 11 | Force Majeure | Standard | Scope of events, notification, mitigation, termination rights |
| 12 | Payment Terms | Standard | Net terms, late fees, taxes, price escalation, disputed invoices |
| 13 | Compliance & Regulatory | Standard | Anti-bribery, sanctions, export control, modern slavery, regulatory approvals |
| 14 | Definitions & Boilerplate | Standard | Entire agreement, severability, waiver, notices, counterparts, amendments |
After analyzing the clauses that are present, check for important clauses that are entirely absent from the contract. A missing clause can be as significant as a problematic one.
For each of the 14 clause categories:
Common missing clause issues:
⟁ CLARIFY — When the severity of a missing clause depends on business context that hasn't been provided, ask rather than guess:
Classify each deviation from the playbook using the three-tier system defined in the Deviation Severity Classification section below. For each deviation:
⟁ CLARIFY — For borderline classifications where reasonable minds could differ, ask the user rather than assuming:
⟁ CLARIFY — Before generating redlines, ask the user about negotiation posture if not already clear from context:
For each YELLOW and RED deviation, generate a redline using the format in Redline Format below. Follow the Redline Best Practices to ensure redlines are specific, balanced, and professionally rationed.
Provide a summary section covering:
⟁ CLARIFY — If the deal context is unclear and it would materially affect the negotiation strategy, ask:
Before delivering the analysis, run the quality checks defined in the Quality Assurance Framework section:
If a Contract Lifecycle Management system is connected via MCP:
If no CLM is connected, skip this step.
Key elements to review:
Common issues:
[JURISDICTION-SPECIFIC] Key variations:
Key elements to review:
Common issues:
[JURISDICTION-SPECIFIC] Key variations:
Key elements to review:
Common issues:
[JURISDICTION-SPECIFIC] Key variations:
Key elements to review:
Common issues:
[JURISDICTION-SPECIFIC] Key variations:
Key elements to review:
Common issues:
[JURISDICTION-SPECIFIC] Key variations:
Key elements to review:
Common issues:
[JURISDICTION-SPECIFIC] Key variations:
For the following clause categories, review presence, reasonableness, and alignment with the playbook. Flag deviations using the same GREEN/YELLOW/RED system.
7. Confidentiality
8. Representations and Warranties
9. Insurance
10. Assignment
11. Force Majeure
12. Payment Terms
13. Compliance and Regulatory
14. Definitions and Boilerplate
The clause aligns with or is better than the organization's standard position. Minor variations that are commercially reasonable and do not materially increase risk.
Examples:
Action: Note for awareness. No negotiation needed.
The clause falls outside the standard position but within a negotiable range. The term is common in the market but not the organization's preference. Requires attention and likely negotiation, but not escalation.
Examples:
Action: Generate specific redline language. Provide fallback position. Estimate business impact of accepting vs. negotiating.
The clause falls outside the acceptable range, triggers a defined escalation criterion, or poses material risk. Requires senior counsel review, outside counsel involvement, or business decision-maker sign-off.
Examples:
Action: Explain the specific risk (with legal basis where possible). Provide market-standard alternative language. Estimate exposure. Recommend escalation path.
For each redline:
**Clause**: [Section reference and clause name]
**Current language**: "[exact quote from the contract]"
**Proposed redline**: "[specific alternative language]"
**Rationale**: [1-2 sentences explaining why, suitable for external sharing]
**Legal basis**: [Relevant statute, regulation, or principle — or "General commercial practice"]
**Priority**: [Must-have / Should-have / Nice-to-have]
**Fallback**: [Alternative position if primary redline is rejected]
When presenting redlines, organize by negotiation priority:
Issues where the organization cannot proceed without resolution:
Issues that materially affect risk but have negotiation room:
Issues that improve the position but can be conceded strategically:
Negotiation strategy: Lead with Tier 1 items. Trade Tier 3 concessions to secure Tier 2 wins. Never concede on Tier 1 without escalation.
Apply structured discipline to every contract review:
PLAN: Identify contract type, user's side, playbook position. Classify complexity (standard / bespoke / high-value). Identify which statutes, regulations, and legal principles are likely to be engaged.
DO: Execute the clause-by-clause analysis and missing clause detection. Generate redlines. Score severity.
CHECK: Run the Citation Quality Gates. For any RED item, run the Self-Interrogation. Verify all statutory references against authoritative sources where possible. Confirm all 14 clause categories addressed.
ACT: Record any new patterns discovered (novel clause structure, emerging market position). Flag heuristics for future reviews. If the review changes your understanding of a standard position, note it for playbook update.
Run these 5 gates silently before delivering any output. If any gate fails, revise before delivering.
| Gate | Rule | Fail Action |
|---|---|---|
| Source | Every legal claim cites a specific statute, regulation, or established principle | Add citation or mark as "[UNVERIFIED — counsel to confirm]" |
| Format | All citations follow a consistent, recognizable format for the jurisdiction | Fix format |
| Currency | Every cited provision checked for amendments or repeal against authoritative sources where accessible | Flag as "[CHECK CURRENCY — may have been amended]" |
| Domain | Analysis stays within the contract's governing law. No assumptions from other jurisdictions leaking in | Remove or flag jurisdictional bleed |
| Confidence | Uncertainty explicitly stated, not hidden. If you are not certain of a legal position, say so | Add confidence qualifier |
For any clause classified as RED, apply this 3-pass self-interrogation before delivering:
Pass 1 — Legal Chain Integrity:
Pass 2 — Completeness:
Pass 3 — Challenge:
If any pass reveals a weakness, revise the analysis before delivery. Mark the audit trail
with self_interrogation: PASS or self_interrogation: REVISED.
For each material clause analysis, assign a confidence level:
| Level | Range | Meaning | Action |
|---|---|---|---|
| Definite | 0.95-1.0 | Settled law, clear statute, no ambiguity | State with confidence |
| High | 0.80-0.94 | Strong authority, minor interpretation questions | State with brief caveat |
| Probable | 0.60-0.79 | Good arguments but reasonable minds could differ | State with explicit reasoning and contra-indicators |
| Possible | 0.40-0.59 | Genuinely uncertain, competing authorities | Flag for counsel review with both sides of the argument |
| Unlikely | 0.0-0.39 | Weak basis, speculative | Do not assert; flag as "[UNCERTAIN — counsel to advise]" |
Every contract review output MUST include a Glass Box audit section at the end. This makes the reasoning traceable and auditable:
glass_box:
contract: "[Contract title and date]"
contract_type: "[SaaS / Professional Services / License / etc.]"
user_side: "[Supplier / Customer / Licensor / Licensee]"
governing_law: "[Jurisdiction identified from the contract]"
playbook_used: "[Playbook name or 'General commercial standards']"
review_basis: "[Organizational playbook / General commercial standards]"
clauses_reviewed: 14
clauses_present: "[number]"
clauses_missing: "[number] — [list]"
legalcode_mcp: "Connected / Not connected"
authority_reference_file: "[path to temp file or 'Not created']"
statutes_consulted:
- "[Statute 1 — VERIFIED (legalcode-mcp) or UNVERIFIED]"
- "[Statute 2 — VERIFIED (legalcode-mcp) or UNVERIFIED]"
case_law_consulted:
- "[Case 1 — VERIFIED (legalcode-mcp) or UNVERIFIED]"
citations_verified: "[number VERIFIED] / [number UNVERIFIED]"
self_interrogation: "PASS / REVISED / NOT APPLICABLE (no RED items)"
confidence: "HIGH / MEDIUM / LOW — [rationale]"
limitations:
- "[Any scope limitations, assumptions, or caveats]"
reviewer: "AI-assisted — requires qualified legal review"
For every contract, identify ALL affected stakeholders — not just the two contracting parties:
| Stakeholder | Role | Affected Clauses | Impact | Action Required |
|---|---|---|---|---|
| [Party A] | [Customer] | [All] | Primary | Sign / Negotiate |
| [Party B] | [Supplier] | [All] | Primary | Sign / Negotiate |
| [Data subjects] | Third party | Data protection | Indirect | DPA required |
| [Sub-processors] | Third party | Data protection, liability | Indirect | Notification rights |
| [Regulator] | Regulator | Data protection, compliance | Compliance | Breach notification |
| [Employees] | Internal | IP, confidentiality | Indirect | Employment contract alignment |
This prevents the common failure of reviewing a contract as a two-party document when it affects multiple stakeholders with different interests and regulatory requirements.
Explicit catalogue of what NOT to do in contract review:
Single-pass analysis — A contract review that reads each clause once, in order, will miss interactions between clauses (e.g., an uncapped indemnity partially mitigated by a limitation clause). Read the whole contract, then analyze.
Citing statutes from memory without verification — Statutes are amended, repealed, and renumbered. Case citations may not exist. If you cannot verify a citation against an authoritative source, mark it [VERIFY] or "[UNVERIFIED — counsel to confirm]." Never present an unverified citation as settled law.
Applying jurisdiction-specific concepts universally — Work-for-hire exists in the US but not England or civil law systems. Kary umowne (penalty clauses) are a Polish concept with no direct common law equivalent. UCTA is English law only. Always check whether a concept applies under the contract's governing law.
Treating indemnities as equivalent to damages claims — In many common law jurisdictions, an indemnity is a primary obligation with different remoteness rules and limitation periods than a damages claim. In civil law jurisdictions, the concept may not be native at all. Analyze accordingly.
Reviewing limitation clauses without checking statutory enforceability — Many jurisdictions impose statutory limits on what liability can be excluded (UCTA in England, unconscionability in the US, mandatory civil code provisions in civil law). Skipping this check is a material omission.
Accepting "consequential loss" at face value — The contract may redefine "consequential" or "indirect" differently from the legal default. Always check the definitions clause.
Ignoring the penalty doctrine for termination fees — Early termination fees and liquidated damages must be assessed against the governing law's penalty doctrine. Don't just flag the amount — assess proportionality.
Redlines without fallback positions — A redline that says "delete this clause" without offering an alternative is a negotiation dead-end. Always provide a fallback.
Importing terminology from the wrong jurisdiction — "Attorney's fees," "punitive damages," "class action waiver," and "jury waiver" have no meaning in many jurisdictions. Don't import them into analysis governed by a legal system where they are irrelevant.
Confidence without evidence — Never state "this clause is unenforceable" without citing the specific statute, regulation, or principle that renders it so. Confident assertions without authority are more dangerous than expressed uncertainty.
Reviewing for one party while assuming the other's interests — The analysis must be from the user's stated side. What protects a vendor harms a customer and vice versa. Confirm the side before analyzing.
Skipping the holistic view — Individual clause analysis is necessary but not sufficient. The contract must also be assessed as a whole: is the overall risk allocation balanced? Do the commercial terms make business sense? Are any provisions internally contradictory?
Ignoring what is missing — Focusing only on clauses that are present and ignoring absent clauses. A contract with no limitation of liability, no DPA, or no termination rights has material gaps that must be flagged.
Treating boilerplate as irrelevant — Entire agreement clauses, severability, notice provisions, and order of precedence can have significant legal consequences. A missing entire agreement clause may allow parol evidence; an incorrect order of precedence can invert the deal terms.
Assuming instead of asking — Silently assuming the user's side, risk tolerance, deal context, or preferred outcome when it would change the analysis. A liability cap that is RED for a risk-averse enterprise client may be YELLOW for a startup that needs the deal. When in doubt, ask — a wrong assumption produces a confidently wrong analysis.
Apply plain-language discipline to all output:
For redline rationales (shared with counterparty's counsel):
For internal analysis:
Quality gates before delivery:
This skill is designed to work with legalcode-mcp as its primary legal research tool. legalcode-mcp is a law, regulatory, and case law database with powerful search across jurisdictions.
With legalcode-mcp connected (preferred):
/tmp/legalcode-review-authority.md)Without legalcode-mcp:
external_tools_used: "None — manual verification required"Structure the final deliverable as:
## Contract Review Summary
**Document**: [contract name/identifier]
**Parties**: [party names and roles]
**Your Side**: [vendor/customer/etc.]
**Governing Law**: [jurisdiction]
**Deadline**: [if provided]
**Review Basis**: [Playbook name / General Commercial Standards]
**Date of Review**: [date]
---
## Key Findings
[Top 3-5 issues with severity flags (RED/YELLOW) and one-line summaries]
---
## Missing Clauses
[List any of the 14 clause categories that are absent, with severity assessment]
---
## Clause-by-Clause Analysis
### [Clause Category] — [GREEN / YELLOW / RED] | Confidence: [level]
**Contract says**: [summary of the provision]
**Standard position**: [playbook or market standard]
**Deviation**: [description of gap]
**Business impact**: [what this means practically]
**Redline** (if YELLOW or RED):
> [Specific proposed language]
[Repeat for each clause category]
---
## Negotiation Strategy
**Tier 1 (Must-Haves)**: [list]
**Tier 2 (Should-Haves)**: [list]
**Tier 3 (Concession Candidates)**: [list]
**Recommended approach**: [sequencing, tone, strategic trades]
---
## Stakeholder Impact Map
[Table per Multi-Stakeholder Mapping section]
---
## Next Steps
[Specific actions to take, with owners and deadlines if applicable]
---
## Glass Box Audit Trail
[YAML block per Glass Box section]
This skill is jurisdiction-agnostic by design. To create a jurisdiction-specific variant:
Existing jurisdiction-specific variants:
imported-uk-legal-plugins--skills-uk-legal-contract-review/SKILL.mdimported-iura-plugins--iura-pl-skills-contract-review/SKILL.mdCreated by Legalcode (2026-02-27). Original synthesis combining playbook-based review methodology, quality assurance frameworks, confidence scoring, Glass Box audit trails, and jurisdiction-agnostic clause analysis into a single contract review skill.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub roberthh-is/legalcode-plugin --plugin legalcode-claude-code