Corporate / Transactional
Contract Review / Clause Extraction
Output: TableModel target: EitherLast evaluated: Jun 18, 2026
hallucination-riskattorney-review-required
How this console works
Law firms have no structured way to author, test, govern, and teach the prompts their teams use — so prompts live in people’s heads and chat histories. This console treats prompts as managed firm assets, with a lifecycle:
- 1Author
Catalogue prompts as versioned assets with guardrail tags and documented edge cases.
Prompt Library
- 2Test
Compare variants A/B on real inputs and score them — manually and with automated checks + an LLM judge.
Test Bench
- 3Prove & Govern
Run golden-case regression tests, and track every documented way the tool can go wrong.
Regression · Guardrails
- 4Teach
Show attorneys how to adapt a prompt safely, without breaking its safety rails.
Playbook
The four prompts below are seed examples that demonstrate the craft and methodology. In a firm deployment this is a living library the practice-innovation team grows and versions over time.
Prompt Library
Prompts managed as versioned assets — each with a defined output template, guardrail tags, documented edge cases, and a failure-mode register. Expand any card to read the full system prompt exactly as deployed, copy it, or open it in the Test Bench.
v1.3
Corporate / M&A
Due Diligence Checklist Generation
Output: ChecklistModel target: EitherLast evaluated: Jun 11, 2026
jurisdiction-sensitivehallucination-risk
Litigation / Firm-wide
Legal Research Summary
Output: MemoModel target: ClaudeLast evaluated: Jun 25, 2026
hallucination-riskattorney-review-requirednever-client-facing
Firm-wide / Intake
Matter Intake Triage
Output: Triage cardModel target: EitherLast evaluated: Jun 20, 2026
confidentialitydata-leakageclient-facing-adjacent