How to Build Claude Skills That Pull Real Case Law
There's a difference between asking Claude to research a legal issue and having Claude actually search a database of real court opinions before it answers you. This is how to build the second one — for free, in under 30 minutes, without writing a single line of code.
Key Takeaways
- •Small and mid-size law firms can compete with Big Law by implementing structured operational systems — without the overhead.
- •Enterprise-level financial management, back office operations, and HR systems drive profitability and scalability.
- •The 1800 Hour Rhythm creates predictable revenue, cash stability, and accountability without attorney burnout.
- •Operational excellence — not lower rates — is your competitive advantage against larger competitors.
What You Need
Claude Cowork (desktop app). The skill system lives here. If you're using Claude in a browser, you can't install skills.
CourtListener — free legal research database run by the Free Law Project. Covers millions of federal and state opinions, PACER dockets, and RECAP documents. No subscription. No Westlaw bill.
That's it.
Step 1: Connect CourtListener to Claude
Open Claude Cowork. In the left sidebar, find Plugins or Connectors — depending on your version, it may be labeled differently. Search for CourtListener in the MCP registry.

Hit Connect. CourtListener requires no API key for read-only access. Once it's connected, Claude can query millions of court opinions directly inside any conversation.
You'll know it worked when you see CourtListener listed as a connected tool. From that point on, Claude can run real searches — not recall from training memory — any time you ask about case law.
Step 2: Understand What a Skill Is
A skill is a set of instructions that tells Claude exactly what to do when a specific type of request comes in.
Without a skill, you write a long prompt every time. With a skill, Claude already knows the workflow — which tools to call, in what order, what to do with the results, and what to flag when something looks wrong.
A skill is a single file: SKILL.md. It lives inside a zip archive with a .skill extension. Claude reads it automatically when the skill is installed and triggered.
Step 3: Write the SKILL.md
Create a new text file named SKILL.md. It must start with YAML frontmatter — three dashes, the metadata, three dashes — followed by the instructions.

Here's the structure:
---
name: case-law-research
description: "Use this skill when an attorney asks for case law research, wants to find relevant precedent, or asks 'what cases support...' or 'how have courts ruled on...'. Searches CourtListener's real opinion database."
---
# Case Law Research Skill
## Steps
### Step 1: Gather Input
Ask the attorney for the legal issue, jurisdiction, and which side of the issue their client is on.
### Step 2: Search CourtListener
Call the CourtListener search tool with:
- type: "o" (opinions)
- q: the legal issue
- court: the relevant circuit or court code
- order_by: "citeCount desc"
- stat_Published: true
- num_results: 10
### Step 3: Read the Key Opinions
For the top 3-4 results by cite count, call read_document using the opinion ID. Read chunk 0 first.
### Step 4: Deliver the Research Brief
Output: controlling doctrine, cases supporting client's position (with citations and CourtListener URLs), cases against (with how courts have distinguished them), recent trend, and what still needs Westlaw verification.
## Rules
- Only cite cases retrieved from CourtListener in this session. Never use recalled training data for case citations.
- Flag every case you couldn't verify.
- Always include the CourtListener URL for every case cited.
The description field is the trigger. Claude reads it to decide when to invoke the skill. Write it like you're describing the exact moment someone would need this — specific phrases attorneys actually use, not abstract categories.
Step 4: Package It as a .skill File
Open Terminal (Mac) or Command Prompt (Windows).
Navigate to the folder where your SKILL.md lives:
cd /path/to/your/folder
Zip it:
zip case-law-research.skill SKILL.md

That's the installable file.
Step 5: Install It in Claude Cowork
In Claude Cowork, go to Settings → Capabilities → Skills. There should be an option to add a skill from a file. Drag in the .skill file or select it from the file picker.
The skill name and description you wrote in the frontmatter will appear in your skills list. From this point on, any time your conversation matches the trigger description, Claude will automatically load and follow the skill's instructions.
The Five Skills I Built (Ready to Install)
If you don't want to build from scratch, here are five production-ready skills covering the most research-heavy legal tasks:
- case-law-research — Searches opinions by cite count, reads key cases, delivers a structured research brief
- opposition-research — Searches federal dockets by attorney or firm name, builds a litigation profile
- brief-authority — Pulls controlling authority and key quoted language for a specific motion standard
- statute-interpreter — Finds cases interpreting a statute and maps each element to how courts have applied it
- legal-risk-assessment — Searches for cases involving similar conduct, calibrates risk as High / Moderate / Low with case support
Each one is a .skill file. Install them the same way as above.
One Important Caveat
CourtListener has strong federal appellate coverage. State trial court coverage is thinner. For state court matters, these skills are a starting point — run the final cite-check on Westlaw or Lexis before anything goes in a filing.
The skill is designed to know this. Every research brief it produces ends with a section called What You Still Need to Verify — specifically because the goal is a tool attorneys can trust, not one that gives them false confidence.
The Jakovenko Group — jakovenko.io
Get the full course on AI legal research and more at LegalAdministrator.ai. If you need a partner to help scale your firm's profitability, check out our Margin Wall Calculator or pick up a copy of The Millionaire Law Firm.
Stop guessing and start grounding your AI in reality.
Want Enterprise Systems Without Big Law Overhead?
We bring AmLaw 200-level operational infrastructure to growth-focused law firms — without the massive overhead.
Related Articles

The Ultimate Guide to AI Ethics for Law Firms: Everything You Need to Succeed
Most law firms are making one of two fatal errors: ignoring AI ethics entirely or overcorrecting by banning the technology. State bar associations aren't banning AI — they're demanding competence. Here is the high-signal breakdown of where the ethics landscape stands in 2026 and how to navigate it without getting disbarred.

Are Law Firm Manuals Dead? How AI Is Rebuilding Firm Operations
Traditional law firm policy manuals are static artifacts that get updated once every three years. In a high-stakes environment where a $280K Margin Wall is breathing down your neck, you cannot afford to have your operational intelligence locked in a static file. AI is turning the manual from a passive document into an active operations layer, reducing Time-to-Competence for new hires by up to 50%.

7 Mistakes Attorneys Make with AI Drafting (and How to Fix Them)
Most attorneys treat LLMs like a magic search engine when they should be treating them like a junior associate who has the entire law library memorized but lacks the context of your specific case. Firms that implement a standardized prompting anatomy see a 28% reduction in total drafting time without any dip in quality or increase in risk.