Guide · Agent skill
How to use the Worker Classification Check agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Worker Classification Check is an agent skill that walks an AI agent through a structured employment-law checklist before you bring someone on as an employee, independent contractor, temp or vendor. It asks how the work will actually run, applies the classification tests for the relevant jurisdiction, and flags where the label you intend does not match the facts. This guide explains what the skill does, how to install it, how to trigger it with plain requests, and what it needs and does not do.
Worker Classification Checkanthropics/claude-for-legal · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the worker-classification ZIP from this page and unzip it so the folder contains SKILL.md.
- 2Place the skill folder. Move the folder into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. Alternatively, upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Set up CLAUDE.md. Create the config at ~/.claude/plugins/config/claude-for-legal/employment-legal/CLAUDE.md with your jurisdictional footprint and escalation table.
- 4Confirm the arrangement is prospective. Check that the worker has not started work. If they have, the skill will stop and route the matter to counsel.
- 5Describe the proposed arrangement. Give the role, duration, location, tools, schedule and pay structure in plain language, or start the slash command and answer the questions.
- 6Review the flagged gaps. Read where the intended label does not match the facts and decide what to change in the structure or operations.
- 7Take the output to counsel. Use the output as a first-pass checklist and get jurisdiction-specific advice before signing.
What the worker-classification skill does
Agent skills are instruction packs that an AI agent loads when a task matches them. This one is for a single question: should the proposed worker be classified the way you plan to classify them? It is designed for arrangements that have not started yet, because the cost of misclassification grows the longer work continues.
The skill asks how the proposed work will actually run, including who sets hours, which tools are used, where the work happens, the degree of control, and whether the worker is exclusive. It then runs the applicable classification tests and reports where the intended structure and the facts diverge.
- Checks employee, IC, temp and vendor structures against the facts you describe
- Asks for all missing details in one block rather than one question at a time
- Stops and routes to counsel if the work has already begun
Install the skill from the ZIP
Download the ZIP for the worker-classification skill from this page, then unzip it. The skill is delivered as a folder containing SKILL.md. Place that folder in one of two locations depending on how you want to use it.
For a personal install available across your projects, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload
Set up the required CLAUDE.md config
The skill relies on a configuration file for the jurisdictions your organisation operates in and for your escalation rules. Without it, the skill cannot identify the correct tests or who to escalate to. The file is expected at the path below, and it should contain the jurisdictional footprint and an escalation table.
The skill also reads any house classification policy and prior classification history that your team has recorded in that file.
~/.claude/plugins/config/claude-for-legal/employment-legal/CLAUDE.mdTrigger the skill with natural requests
You can invoke the skill with its slash command or simply describe the arrangement. The skill's description lists phrases such as "we want to bring on a contractor", "is this a vendor or a temp", and "how should we classify this person", so a plain-language request should load it.
If you give the details up front, the skill extracts what it can and asks only about the gaps. If you give nothing, it will ask for the details.
- "We want to bring on a contractor for a six-month analytics project."
- "Is this vendor or a temp? They would sit in our team and take daily direction."
- "How should we classify this person?" followed by a description of the role
/employment-legal:worker-classificationExamples of prompts and what to expect
A complete prompt includes the role, the duration, the location, the tools, and how the person is paid. The following example gives the skill enough to work with.
Describe the arrangement as the worker would experience it, not only the label you plan to use. The skill tests the facts, so a contract that says "independent" while the person works fixed hours on your tools will be flagged.
- Role and duration: a data scientist for six months
- Location and tools: your office and your equipment
- Structure: embedded in a team, paid by project fee or salary
/employment-legal:worker-classification
We want to bring on a data scientist for 6 months, working out of our
SF office, using our tools, embedded in our analytics team.Safety, scope and what it needs
The skill is prospective only. If you say the work has already started, it stops and recommends escalation to counsel instead of analysing the arrangement. It will not produce a remediation plan for existing workers. If you still want forward-looking planning after that warning, you must explicitly say "proceed anyway", and the output will carry a scope-mismatch banner.
The SKILL.md states that it teaches the reasoning pattern and does not itself state the law. Jurisdiction-specific tests must come from current research and counsel. The skill does not run scripts and does not use the network.
- Requires Claude Code or a compatible agent with the claude-for-legal employment-legal plugin installed
- Requires a configured CLAUDE.md with jurisdictional footprint and escalation table
- Runs scripts: no
- Uses network: no
- Licence: Apache-2.0
Tips for better results
The more specific your description, the more useful the check. Focus on how the work runs in practice. Note whether the worker is exclusive to you, whether they set their own hours, and whether they use their own tools. Then treat the output as a structured first pass, not a final answer.
- Describe the actual day-to-day work, not just the job title
- Say whether the work is core to the business or peripheral
- State whether the project has a defined end date
- Keep your CLAUDE.md jurisdictions current so the right tests apply
What people use it for
- An HR team planning a new contractor engagement checks the structure before posting the role.
- A manager deciding whether a proposed role should be an employee or a contractor.
- An employment lawyer preparing a structured first-pass checklist on a prospective arrangement.
- A legal team reviewing whether a recruiter or vendor arrangement fits the intended label before signing.
Frequently asked questions
How do I install the worker-classification skill?
Download the ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
Is the worker-classification skill free?
The source material does not state pricing or licence fees for the skill. It is listed under the Apache-2.0 licence, so check the licence text in the repository for the terms that apply.
Does the skill give legal advice?
No. The SKILL.md says it teaches the reasoning pattern and does not state the law. Jurisdiction-specific tests must come from current research and counsel.
Can I use it on a worker who is already working for us?
Not as a classification check. The skill stops when the work has already started and routes the matter to counsel. It only continues if you explicitly say "proceed anyway", and then the output carries a scope-mismatch warning.
What do I need to run it?
You need Claude Code or a compatible agent with the claude-for-legal employment-legal plugin installed, plus a configured CLAUDE.md at ~/.claude/plugins/config/claude-for-legal/employment-legal/ containing your jurisdictional footprint and escalation table.
Does the skill run scripts or access the internet?
According to the listing, it runs no scripts and uses no network. It works from the SKILL.md instructions and your CLAUDE.md config.
Why does it need a CLAUDE.md file?
The skill relies on that file for your jurisdictional footprint, classification history, house policy and escalation rules. Without it, the skill cannot pick the correct tests or the right person to escalate to.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are planning a new contractor, temp, vendor or employee engagement and want a structured first pass before signing, especially if you have a configured employment-legal plugin. Pick something else if the worker is already engaged, if you need a remediation plan for past work, or if you are outside the jurisdictions your CLAUDE.md covers, and in those cases talk to employment counsel directly.
Download the skill