How to use the Legal Research Roadmap Starter agent skill

What is an agent skill and what does research-start make the agent do?

An agent skill is a folder containing a SKILL.md file with instructions that an agent loads when a request matches the skill's description. The research-start skill is designed for the first phase of legal research, when a student is working out what to look for rather than what the answer is.

When triggered, the agent loads the clinic's configuration file for jurisdiction and practice area, and reads the clinic's own seed documents first if any match the issue. It then frames the question specifically, builds a roadmap, and labels everything as a lead to verify. If the student has already uploaded research, the agent synthesizes it and identifies gaps.

• Statutory starting points, each marked UNVERIFIED for currency and applicability
• Case law areas to investigate, not specific cases
• Regulatory or administrative sources where relevant
• Secondary sources for orientation, not for citation
• Westlaw and CourtListener search terms to start from
• Flags on anything uncertain, and source tags on any suggested citation

How do you install the research-start skill?

Download the ZIP for the skill from the directory page and unzip it. The skill is a single SKILL.md file, so you can place it in either of two locations depending on how you want to use it.

For a personal skill available across projects, unzip it into ~/.claude/skills/. For a skill that only applies to one project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps rather than Claude Code, you can upload the skill instead under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Set up the clinic configuration file before you run it

The skill reads a clinic configuration file at ~/.claude/plugins/config/claude-for-legal/legal-clinic/CLAUDE.md. The skill uses this file for the jurisdiction (state) and practice areas, and looks for a Seed documents section listing the clinic's uploaded materials.

Without this file the skill cannot load your jurisdiction or seed documents, so it is not suitable for general use outside a clinic configured this way.

How do you trigger the skill with a natural request?

The skill's description says it applies when a student asks where to start researching, wants a research roadmap for an issue, or needs gaps identified in existing research. You can invoke it with its slash command and a short issue description in quotes, or describe what you need in plain language that matches those situations.

Be specific about the issue. The skill narrows broad questions, and it will ask you which sub-question to take first if your question contains several.

• Where should I start researching whether a broken heater is a habitability defense to nonpayment eviction?
• Give me a research roadmap for an unpaid wage claim under my state's wage law.
• Here is my existing research memo. What gaps do I still have on this issue?

What does the output look like and how should you read it?

Output opens with a header stating that the content is leads, not authorities. Seed documents from the clinic, if any match, appear first. Every statute, case law area and search term is a starting point you verify yourself.

Citations carry source tags showing where they came from, such as Westlaw, CourtListener, web search (verify), model knowledge (verify) or user provided. Keep these tags: they tell you which leads are raw research and which are guesses to check first.

• Read seed documents before anything else, if the skill found matches
• Treat every statute as unverified and confirm it is current
• Check any tag marked verify against a primary source before relying on it
• If a search returns few results, the skill will say so and offer options rather than invent citations

What requirements and safety notes apply?

The skill needs Claude Code or a SKILL.md-compatible agent and the clinic configuration file described above. Westlaw or CourtListener access is needed only if you want the agent to run searches; the skill's files reference these tools, but no API keys are specified in the source.

The skill does not run scripts. It does use the network, because it can query research tools and, if you approve, search the web. Review what the agent proposes to query before approving.

• Requires: Claude Code or a SKILL.md-compatible agent
• Requires: clinic configuration file at the path above
• Optional: Westlaw or CourtListener access for live searches
• Runs scripts: no
• Uses network: yes
• Licence: Apache-2.0

Tips for getting useful roadmaps

Narrow the question before you start. A question naming a state, a defense and a specific fact produces a far more useful roadmap than a general topic.

Upload any research you already have. The skill's gap analysis depends on seeing what you have found, and it can flag a case you cited that needs checking in a citator.

• Name the state and the specific fact pattern in your request
• Ask one sub-question at a time when the skill splits your issue
• Run the suggested search terms yourself and refine them from the results
• Take the roadmap to your supervisor to confirm which leads matter