18 legal skills for Claude: a legal AI operating system

Eighteen standalone Claude skills for legal work, grouped into six practical practice domains. Each skill is a small, uploadable workflow that teaches Claude how to perform one high-value legal task with a repeatable, lawyer-grade method.

What this is

Together, the skills operate like a legal AI operating system for the full lifecycle of legal work: review and draft the contract, run the diligence, keep the company compliant, handle the dispute, answer the hard question, and communicate the conclusion so the client actually understands it. Built for in-house counsel, law firm associates, compliance officers, and legal ops teams who want Claude to work closer to an experienced practitioner — structured before verbose, evidence-tied before confident, and honest about uncertainty before persuasive.

Mechanically, each skill is one folder with a single SKILL.md file. Install just the ones you need, or load the full set. All skills are jurisdiction-neutral: they encode method, not local law, and explicitly flag where local-counsel confirmation is required.

What the skills produce

These are real outputs from the skills running in Claude — synthetic, anonymized data, no product interface. Click any one to read it at full size.

How to choose a legal skill

Start from the need, not the skill. Each row links to the full description below.

The 18 skills you get

The collection is grouped into six practice domains. They run in the order legal work actually flows — the deal first, the dispute if it goes wrong, the communication always — but every skill is also useful on its own.

Domain 1 · Contracts & Transactions (4 skills)

The daily bread of legal work. Use these when a contract must be reviewed, drafted, compared, or negotiated — and the output has to be positional, not neutral: risk always depends on which side you're on.

Contract Review (contract-review)

Use when: a contract or counterparty draft needs a risk assessment before signing.

What it does: runs a party-specific review — maps the document, tests every operative clause, checks the standard protection set (liability caps, indemnities, termination, IP, data), and hunts for what's missing, the most common source of loss. Findings are ranked Critical → Low.

Output: an executive summary with a sign/negotiate/walk recommendation, a severity-ranked risk table, and — the differentiator — proposed redline language for every finding, not just a description of the problem.

Contract Drafting (contract-drafting)

Use when: an agreement, amendment, or clause must be built from a term sheet, an email chain, or intent.

What it does: confirms the clause skeleton before drafting, drafts economics first, allocates risk deliberately for your side, then runs a consistency pass on defined terms and cross-references. Every assumed business term is marked [CONFIRM: ...] instead of silently decided.

Output: a complete, signable draft — no "[insert standard indemnity]" placeholders — plus drafting notes on the choices made and the clauses the counterparty will push back on.

Contract Comparison (contract-comparison)

Use when: a counterparty returned a marked-up (or silently edited) draft, or a contract must be checked against your standard template.

What it does: detects every change including the quiet ones — altered defined terms, moved clauses, changed cross-references — and classifies each as cosmetic, clarifying, or substantive. Traces interaction effects: a one-word edit in definitions can rewrite ten other clauses.

Output: a change log with an effect-on-you column and a per-change response (accept / modify / reject-and-restore with counter-language), plus a ready-to-adapt response email.

Negotiation Playbook (negotiation-playbook)

Use when: you're walking into a negotiation, or building a reusable playbook for a recurring contract type.

What it does: builds a three-level position for every contested clause — opening, conditional fallback ("we accept X if they accept Y"), and red line with escalation trigger — each armed with the argument to say aloud and the rebuttal to the likely counter.

Output: a priority map (must-win / tradeable / give-away), clause-by-clause positions, a recommended negotiation sequence, and a prepared response to the counterparty's most likely package offer.

Domain 2 · Due Diligence & Corporate (3 skills)

Use these when the question is what you're buying, who is authorized, or how the deal should be built.

Legal Due Diligence (legal-due-diligence)

Use when: a target, investment, or major counterparty must be checked before committing.

What it does: works in three modes — builds the scoped DD request list going in, reviews data room documents against the standard risk set (change-of-control triggers, consent requirements, unusual liabilities), and consolidates findings into a red flags report. Every finding is classified by deal impact and mapped to a remedy: condition precedent, price adjustment, indemnity, or walk.

Output: a red flags report with quantified exposure, a consents-and-approvals checklist, and — critically — an explicit scope-limitations section, because undisclosed scope gaps are how DD reports go wrong.

Corporate Governance Review (corporate-governance-review)

Use when: you need to know who is authorized to sign and what approvals a transaction requires.

What it does: builds an authority map from the charter, bylaws, and shareholders' agreement; tests the transaction against value thresholds, reserved matters, and veto rights; and runs the interested-party check even when not asked — related-party issues invalidate deals more often than value thresholds do.

Output: an answer-first conclusion (approval required: yes/no/depends), a sourced authority map, an executable step-by-step approval route, and draft resolutions or consents on request.

Deal Structuring Memo (deal-structuring-memo)

Use when: choosing between structures — share vs. asset deal, JV forms, reorganizations — or sanity-checking a proposed one.

What it does: generates 2–4 genuinely different structures (not one option plus two strawmen), compares them on a consistent criteria set — execution risk, consents, liability inheritance, migration of contracts and employees, timing, reversibility — and defends the recommendation against the strongest alternative.

Output: a comparison table, a recommendation that names its own costs, a sequenced implementation step plan with the point of no return identified, and a specialist-referral list for tax and foreign-law items.

Domain 3 · Compliance & Risk (3 skills)

Use these to answer "are we compliant," write the rules, and absorb regulatory change without drowning in it.

Compliance Gap Assessment (compliance-gap-assessment)

Use when: readiness against a regulation or standard must be assessed — for a regulator, an auditor, or an enterprise customer.

What it does: confirms applicability first (which obligations actually apply and which don't), decomposes the regulation into testable requirements, and assesses each with evidence discipline: no evidence means "cannot assess," never optimism. Gaps are rated by enforcement likelihood, sanction severity, and auditor visibility.

Output: a summary scorecard, a full gap register, a phased remediation roadmap where every item produces an auditable artifact, and a targeted evidence-request list.

Policy Drafting (policy-drafting)

Use when: an internal policy, SOP, or code of conduct must be written or updated.

What it does: writes rules as testable obligations — actor, action, trigger, deadline — and checks every rule against organizational reality, because a rule the company won't enforce is a liability in an investigation, not a protection. Includes the operating layer: owners, exceptions process, reporting lines, review cycle.

Output: a complete numbered policy with [DECIDE: ...] brackets for organizational choices, plus implementation notes listing what must exist for the policy to be true on day one.

Regulatory Change Tracker (regulatory-change-tracker)

Use when: a new law, amendment, or guidance has landed and leadership asks "does this affect us and what do we do."

What it does: establishes the exact delta (old rule → new rule, with precise effective and transition dates), filters for applicability to kill the noise early, translates each applicable change into affected processes and documents, and states open interpretation questions with a conservative reading, an aggressive reading, and a recommended position.

Output: a provision-by-provision impact table, a dated action plan tied to effective dates, and a one-paragraph plain-language summary for management.

Domain 4 · Disputes & Litigation (3 skills)

Use these when the relationship broke: decide whether to fight, plan the fight, and fire the first formal shot properly.

Case Assessment (case-assessment)

Use when: sue, defend, settle, or walk away — before money is spent.

What it does: builds the claim architecture (elements → evidence, naming the element with no evidence behind it), steelmans the opponent's best case, estimates outcomes in honest probability bands rather than fake precision, and does the economics — including enforceability, because winning against an insolvent defendant is losing slowly.

Output: a bottom-line recommendation with confidence level, outcome scenarios, the net-expected-value logic, and immediate next steps: evidence to secure, limitation deadlines, preservation duties.

Litigation Strategy (litigation-strategy)

Use when: proceedings are imminent or underway and the case needs a campaign plan, not just pleadings.

What it does: fixes a one-paragraph theory of the case that every move must serve, builds the evidence plan backward from legal elements, sequences procedural moves with their cost and signaling effects, and runs a settlement track in parallel — naming the leverage moments, usually right after a procedural win.

Output: theory of the case, gap-mapped evidence plan, numbered procedural sequence, settlement corridor, pre-decided decision gates ("if the ruling goes against us → ..."), and one consolidated deadline list.

Demand Letter (demand-letter)

Use when: a formal pre-action letter must be sent — payment demand, cease-and-desist, breach notice.

What it does: checks contractual and statutory notice formalities first (a demand that fails formal requirements can cost the case), then drafts for two audiences at once: the counterparty's decision-maker and a future judge. Nothing unprovable, no admissions, no threats the client won't execute, rights reserved.

Output: a ready-for-letterhead letter with calculation annex, plus a cover note: how to send it, to whom, by when, and what to do if there's no response.

Domain 5 · Research & Analysis (2 skills)

Use these when the question is what the law actually says — and the answer has to be traceable, not remembered.

Legal Research Memo (legal-research-memo)

Use when: a substantive legal question needs a reasoned written answer.

What it does: reframes the question precisely, analyzes rule → application → counter-arguments per sub-question, and enforces an absolute citation discipline: verified sources or an explicit [VERIFY: ...] flag — never a remembered citation presented as authority. Every conclusion carries a confidence level: settled / likely / arguable / unsettled.

Output: a conclusion-first memo — a reader who stops after the short answer is still correctly informed — with risks, open points, a recommendation, and a source list that separates verified from flagged.

Legislation Summary (legislation-summary)

Use when: a statute or regulation must be understood in ten minutes instead of three hours.

What it does: works from the actual text, fixes the instrument's exact status and dates (summarizing a superseded version is the classic failure), and extracts the operative core organized by addressee: who must do what, by when, or else what. Adds the interpretation layer — known ambiguities and common misreadings — that separates a summary from a table of contents.

Output: layered depth (one paragraph → one page → full map), a who-must-do-what table with source articles, an enforcement section, and a watch-outs list.

Domain 6 · Client & Business Communication (3 skills)

Use these because legal work only counts when the client understands it. Answer first, jargon last, bad news straight.

Client Advice Letter (client-advice-letter)

Use when: legal conclusions must reach a client or business stakeholder in writing.

What it does: leads with the answer in the first paragraph, translates every legal concept into its practical consequence, delivers bad news straight with an immediate pivot to options, and makes confidence levels visible — clients make different decisions at different certainty levels and can't calibrate what isn't shown.

Output: a ready-to-send letter or email ending in a decision frame: recommended action, alternatives with trade-offs, what's needed from the client, and the deadline logic.

Legal Risk Report (legal-risk-report)

Use when: management or the board needs the consolidated legal exposure picture.

What it does: consolidates by risk theme rather than matter-by-matter diary, quantifies with discipline (exposure ranges and probability bands, with reasons where numbers are impossible), shows movement against the prior period, and ends with decisions requested — because a report without asks trains the board to skim.

Output: a one-page summary that survives being the only page read, a themed risk register with owners and trend arrows, a movement log, and numbered decision requests.

Plain Language Summary (plain-language-summary)

Use when: a contract, judgment, or notice must be understood by non-lawyers — employees, users, the business.

What it does: filters by consequence to this reader, translates by outcome rather than synonym ("indemnification" → "if X happens because of you, you pay for the damage"), and keeps the sharp edges sharp: deadlines, penalties, auto-renewals, and one-sided rights get more emphasis in the plain version, never less. Precision losses are flagged, not hidden.

Output: a structure built around the reader's questions — what do I get, what do I owe, what can go wrong — plus a loud "watch out" list and a note that the original prevails.

Setup guide

Step 1 — Download the skills pack. The zip contains all 18 skills, one folder per skill, each with a single SKILL.md file. Unzip it anywhere. Keep the whole set, or pull out just the folders you want.

Step 2 — Create a Claude Project. Go to claude.ai, open Projects, click Create Project. Name it something like "Legal Assistant" so you can reuse it across matters. Alternatively: upload the folders as Skills under Settings → Capabilities, or drop them into .claude/skills/ for Claude Code.

Step 3 — Add the skills as Project Knowledge. Inside the project, open Project Knowledge, click Add Content, and upload the SKILL.md files. One skill, an entire domain, or all 18 — Claude references them automatically in every conversation inside that project.

Step 4 — Start using the skills. Open a conversation, attach your document, and name the skill you want. Referring to the skill by name ("use the contract-review skill") points Claude at the right method instead of leaving it to guess.

Example prompts

• "Use the contract-review skill on the attached MSA. We are the customer; governing law is English."
• "Use the legal-due-diligence skill to build a phase-1 request list for a share acquisition of a logistics company."
• "Use the case-assessment skill: our distributor stopped paying, contract and correspondence attached."
• "Use the plain-language-summary skill to explain these terms of service to our support team."
• "Run the regulatory-change-tracker skill on the attached amendment — we're a payment institution."

Usage tips

• Name the skill in your prompt. "Use the case-assessment skill" beats hoping Claude picks the right one.
• Give the skill its inputs. Each SKILL.md lists what Claude needs (your party, jurisdiction, goal). Providing these up front skips a round of clarifying questions.
• Chain skills like a workflow. case-assessment → litigation-strategy → demand-letter. Or contract-review → negotiation-playbook → contract-comparison on the counterparty's redline.
• Attach the documents. These skills are built to work on your real contracts, policies, and case files — paste or upload them.

The quality bar

Every skill pushes Claude toward the standard a supervising partner would hold the work to before it leaves the building:

Where the skills stop

These skills are good. They're also honest about what they are: instructions. They teach Claude a method — they don't give it your record.

• They can't open what you don't paste. Every skill works on the documents in one conversation. Your DMS, the data room, the 400-page bundle — out of reach.
• They cite what they can see. [VERIFY: ...] is the right behaviour, but it's still a flag on your desk, not a resolved source.
• They don't remember the matter. Each conversation starts from zero. The clause you negotiated in March isn't there in July.
• They leave no audit trail. Nothing records which document supported which conclusion — the thing you need when someone asks six months later.

Christie is built for exactly that gap: your files, indexed; every answer cited to the exact page; the matter, not the conversation.

These skills are drafting and analysis aids for legal professionals. They do not constitute legal advice, and outputs must be reviewed by a qualified lawyer in the relevant jurisdiction before being relied on or sent. Verify all citations against current primary sources. Check your organization's confidentiality, privilege, and professional-conduct rules before using AI tools on client matters.