What this skill does
Judgment First separates what AI can determine on its own from what requires your judgment as a lawyer - and makes sure the AI never silently crosses that line.
Every legal task has two layers. The intelligence layer: pattern recognition, drafting, lookup, summarizing, precedent mapping. AI handles this well. The judgment layer: decisions that depend on your client's risk appetite, your read of the counterparty's leverage, local practice norms, or choices where reasonable lawyers disagree. AI must not substitute here.
Before producing any legal output, this skill identifies all the judgment-layer decisions in your document or task and asks you about them explicitly. Once you've answered, it produces a Judgment Call Memo - a structured record of what was AI-determined versus lawyer-decided - and AI Work Specs: a scoped instruction set for the rest of the conversation, where every AI task is constrained by the answers you just gave.
When to use it
- Opening any matter where the AI's framing assumptions could quietly skew the analysis
- Reviewing a contract where risk tolerance, leverage, or client priorities must drive the output
- Any situation where you need to be able to show a client, partner, or regulator what judgment calls were made and by whom
- Onboarding a matter with a colleague - the memo becomes a shared strategic briefing
What you get
First, judgment questions - all of them, upfront, with a brief explanation of why each one is yours to answer rather than the AI's. Then, once you've responded:
A Judgment Call Memo with two sections: what the AI determined autonomously (the intelligence layer), and what you decided (the judgment layer), with your answers recorded clearly. Any unanswered questions are flagged with the assumption made.
AI Work Specs: a numbered list of tasks for the rest of the conversation, each one anchored to a specific judgment call you made - so downstream AI work reflects your strategy, not the AI's defaults.
Example prompt inputs
- "Review this SPA - I need to know what judgment calls to make before you start the analysis"
- "We're advising the seller in this M&A transaction. Run Judgment First before drafting any risk memo"
- "Open Judgment First on this employment contract before you redline it"
- "I want a record of the strategic framing before the AI drafts the settlement letter"
Why this matters
AI tools in legal work fail quietly. They make risk-tolerance calls, drafting choices, and strategic framings that look like analysis but encode assumptions the lawyer never reviewed. Judgment First makes those assumptions visible before they become output - and creates a record that the lawyer, not the AI, made the calls that mattered.