--- title: "How the Anthropic Legal Plugin Works (and How to Use It Without Claude Code)" date: "2026-02-02" excerpt: "Anthropic open-sourced a legal plugin with NDA triage and risk classification skills. Here's what the skills actually check, how they work, and how to access them via email without any coding." author: "Steven Obiajulu" tags: ["Legal Tech", "AI", "Anthropic", "Contract Review", "NDA", "Legal Plugin"] readTime: "5 min read" image: "assets/images/blog/anthropic-legal-plugin-how-it-works.png" lastModified: "2026-02-02" --- On January 30, 2026, Anthropic released [Cowork plugins](https://github.com/anthropics/knowledge-work-plugins) — a set of open-source, structured skills for knowledge work. Among them: an official legal plugin. This isn't a chatbot. These are structured workflows that produce consistent, auditable output. The code is public under the Apache-2.0 license. Here's what the legal plugin actually does, what the skills check, and how to use them without running a CLI. ## What Anthropic Released The legal plugin is part of Anthropic's broader Cowork feature. It includes specific "skills" — structured workflows, not free-form prompts: - **NDA Triage Skill**: A checklist-based workflow for reviewing non-disclosure agreements - **Legal Risk Classification Skill**: A severity-by-likelihood framework for contract risk factors The repository is at [github.com/anthropics/knowledge-work-plugins](https://github.com/anthropics/knowledge-work-plugins). Coverage from [Artificial Lawyer](https://www.artificiallawyer.com/2026/02/02/anthropic-moves-into-legal-tech/) and [TechCrunch](https://techcrunch.com/2026/01/30/anthropic-brings-agentic-plugins-to-cowork/) provides additional context. ## What the NDA Triage Skill Checks The NDA triage skill walks through a structured checklist. Each item receives a status: **Standard**, **Needs Attention**, or **Missing**. | Category | What It Checks | |----------|----------------| | **Obligations** | Mutual vs. one-way confidentiality | | **Definitions** | Scope of "confidential information" | | **Carveouts** | Public info, prior knowledge, independent development | | **Term** | Duration of agreement and survival period | | **Termination** | Standard vs. non-standard termination rights | | **Return/Destruction** | Obligations at agreement end | The output is structured — same headings, same format, every time. This makes it useful for comparing across deals and for teams that need consistent first-pass review. ## What the Legal Risk Classification Does The risk classification skill uses a severity-by-likelihood matrix: - **Severity**: The potential impact if the risk is realized - **Likelihood**: The probability of the risk occurring It doesn't tell you what to do. It surfaces factors for lawyer review in a consistent format. Think of it as structured triage, not legal advice. ## How to Use It via Email The Anthropic version runs in Claude Code (a CLI tool) or Claude Cowork. Both require some technical setup. We built an email interface for the same open-source playbooks. Forward a contract to [junior@usejunior.com](mailto:junior@usejunior.com) and receive structured analysis in your inbox: 1. **Forward** any contract to junior@usejunior.com 2. **Receive** a structured triage checklist 3. **Review** the standard vs. non-standard flags with your team No portal login. No software installation. No dashboard to learn. For more details on the email workflow, see [Junior AI Agent](/products/junior-ai-agent.md). ## Our Contribution: PR #6 We submitted [Pull Request #6](https://github.com/anthropics/knowledge-work-plugins/pull/6) to the Anthropic repository. The contribution addresses unauthorized practice of law (UPL) risks by refining terminology: - Changed "assess risk" to "classify risk factors" - Ensured output stays focused on data classification rather than legal advice - Added whistleblower carveout alignment These are subtle but important distinctions for legal defensibility. The PR is public — you can read the changes yourself and contribute your own improvements to the open-source framework. ## Try It Forward a contract to [junior@usejunior.com](mailto:junior@usejunior.com) to see the structured output for yourself. Or [request a walkthrough](/contact.md) to see it in action. ## Related UseJunior Resources - [Junior AI Agent](/products/junior-ai-agent.md) for capability definitions and review guardrails - [Frequently Asked Questions](/faq.md) for workflow and onboarding details - [Security overview](/security.md) for encryption and access controls --- Disclaimer: UseJunior is a productivity tool for legal professionals. We do not provide legal advice. All AI-generated outputs must be reviewed by a licensed attorney. The author practiced law for six years at Ropes & Gray LLP prior to co-founding UseJunior. This implementation is based on the open-source legal plugin content from Anthropic's knowledge-work-plugins repository (Apache-2.0 License). --- References 1. "Anthropic Cowork Plugins Repository," GitHub, accessed February 2, 2026, [https://github.com/anthropics/knowledge-work-plugins](https://github.com/anthropics/knowledge-work-plugins) 2. "Anthropic Moves Into Legal Tech," Artificial Lawyer, February 2, 2026, [https://www.artificiallawyer.com/2026/02/02/anthropic-moves-into-legal-tech/](https://www.artificiallawyer.com/2026/02/02/anthropic-moves-into-legal-tech/) 3. "Anthropic Brings Agentic Plugins to Cowork," TechCrunch, January 30, 2026, [https://techcrunch.com/2026/01/30/anthropic-brings-agentic-plugins-to-cowork/](https://techcrunch.com/2026/01/30/anthropic-brings-agentic-plugins-to-cowork/) 4. "UseJunior PR #6: UPL Safety Tweaks," GitHub, [https://github.com/anthropics/knowledge-work-plugins/pull/6](https://github.com/anthropics/knowledge-work-plugins/pull/6)