Free · no signup

The working toolkit

Most of the prompt library, the model-routing cheat sheet we teach in Week 2, the habits that separate useful output from confident nonsense, and the vocabulary to go with it. Take all of it.

The workWhat to reach forWhy
Discovery review, income analysis, deficiency lettersDeep reasoning modelMulti-document comparison where a missed gap is the whole error.
Pleadings, correspondence, client updatesGeneral drafting modelTone and structure matter more than long-chain reasoning.
Transcripts, status lookups, file renamingFast, cheap modelHigh volume, low judgment. Paying for reasoning here is waste.
Anything touching a retainer or signed documentDeep reasoning + flag-only instructionSilent clause removal is the real risk, not typos.

Pick the model before you pick the prompt

Deep reasoning for income analysis and discovery review. General drafting for pleadings and correspondence. Fast and cheap for transcripts and status lookups. Choosing wrong is the #1 reason output feels mediocre.

Name files before you ask for analysis

A disclosure folder renamed to [date]_[document type]_[party] cuts hallucinated citations dramatically. Structure first, questions second.

Make the model flag instead of guess

Add 'flag anything you can't confidently categorize instead of guessing' to every scan prompt. That one clause turns a liability into a checklist.

Project instructions outlive prompts

Put your role, tone, formatting standards, and the never-do-this-without-asking rules in project instructions. You stop re-explaining your firm every morning.

Never let a clause disappear quietly

Any drafting skill that touches a retainer or pleading should be told to flag removals rather than make them. Silent deletions are the real risk, not typos.

Audit access before you scale it

List every folder the AI can reach and whether access is inherited or explicit. Do it before you add staff, not after.

12 prompts, open.

Case Type & Life Cycle Scan (Cowork)

Build a life-cycle map for a matter type from the files you already have.

Scan every [matter type] file in this folder. For each matter, identify: (1) what stage of the case it's in, (2) what documents exist at each stage, (3) what's typically missing at this stage based on the other matters here. Then build a life-cycle map for this matter type: list every stage from intake to closing, and under each stage list the documents/actions that occur. Flag any stage where documents or steps are inconsistent across matters.

Template & SOP Inventory

Find every clause your firm actually uses, and how often.

Scan our firm's templates and SOP folder for [matter type]. List every unique clause or provision that appears across all versions of this document type. Group similar clauses together and flag which ones appear in most versions vs. only a few.

Project Instructions Generator

Write the standing instructions for a matter-type project.

Write project instructions for a Claude project handling [client/matter type]. Include: my role (attorney/mediator/paralegal), tone requirements, what Claude should never do without asking, and the formatting standards our firm uses (font, spacing, citation style).

Model Selection Self-Check

Pick the right model before you start the task.

Before starting this task, tell me: is this better suited to deep reasoning (income analysis, complex discovery review, building a new skill), general drafting (pleadings, retainers, correspondence), or a fast/cheap lookup (reading a transcript, a quick status question)? Recommend which model I should be using.

Discovery Folder Organization (Cowork)

Rename and sort a mandatory disclosure folder without guessing.

Point Cowork at this mandatory disclosure folder. Rename every file using the convention [date]_[document type]_[party]. Flag any file you can't confidently categorize instead of guessing.

Discovery Tracker Build

Turn the folder into a Rule 12.285 tracker with gaps flagged.

Scan this mandatory disclosure folder and build a tracker with columns: document type, date received, account holder (if financial), last four of account number, status (received/missing/incomplete). Flag every item required by Florida Rule 12.285 that isn't present.

Compare FA to MD Skill

Cross-check the sworn affidavit against the production, line by line.

Compare the sworn Financial Affidavit in [file] against the mandatory disclosure documents in [folder]. For every income, asset, and expense line on the affidavit, identify whether the disclosure documents support, contradict, or fail to address that line. Produce a discrepancy memo listing each conflict with the specific source document and page.

Income Assessment → Deficiency Letter

Go from raw financial records to a sendable deficiency letter.

Using the attached 1099s/bank statements/pay stubs for [party], calculate actual income over the trailing 6 months. Compare this to the income reported on their Financial Affidavit. If there's a discrepancy, draft a deficiency letter to opposing counsel identifying the specific undisclosed deposits/income, with dates and amounts.

Billing Draft Assist

Clean up the month's time entries before the invoice goes out.

Review this month's time entries for [client]. Flag any entries missing detail, any block-billed entries over 1 hour, and draft the narrative language for the invoice cover memo.

Personal Voice Skill Builder

Turn your own writing into a reusable voice skill.

Here are 5 examples of my own writing [attach client letters/blog posts/emails]. Analyze my voice: sentence length, word choice, tone, how formal or casual I am, phrases I repeat. Then write a skill that instructs Claude to draft in this voice going forward, with examples included.

Internal AI Use Policy Draft

A one-page policy a non-technical staff member can follow.

Draft a one-page internal AI use policy for our firm covering: what tasks staff can use Claude for without approval, what requires attorney sign-off, data handling rules, and what's prohibited. Keep it in plain language a non-technical staff member can follow.

Permissions Audit

Confirm what Claude can actually reach before you scale access.

List every SharePoint/OneDrive folder Claude currently has access to. For each, confirm whether access is inherited or explicitly granted, and flag any folder where access looks broader than necessary.

Project

A persistent workspace with its own instructions and files. Your firm's standards live here, not in each chat.

Project instructions

The standing rules — role, tone, formatting, never-do-this-without-asking. They outlive any single prompt.

Skill

A reusable, named procedure the model can invoke. A prompt you stopped retyping.

Connector

A live link to a system of record (files, mail, calendar). Powerful, and the first thing to audit.

Scheduled agent

A task that runs on a clock instead of waiting for you to start it.

Flag-only instruction

A clause telling the model to surface uncertainty rather than resolve it. The single cheapest safety control you have.

  • Firm Analytics Scan (Claude in Chrome)Week 4
  • Retainer Skill BuildWeek 5
  • Week 1 guide · Case Life Cycle
  • Week 2 guide · Claude & Cowork Architecture
  • Week 3 guide · Discovery & Financial Workflows
  • Week 4 guide · Firm Analytics, Billing & Time
  • Week 5 guide · Build Your Own Skills
  • Week 6 guide · Ethics, Security, Rollout & Capstone

Registrants also get the six click-through week guides, the client deliverables library, the build library, group chat, weekly office hours, and two private 1:1 hours.

Seats are limited — 5 per track · 10 total

$1,700 per attorney. Track selection happens in your registrant portal after you enroll, first come, first served.

Enroll now