Turn a Vague Research Need Into a Usable Prompt
Learn the difference between a question that invites a confident guess and a question that invites verifiable research.
You will be able to write a prompt that gives the assistant jurisdiction, posture, task, and verification expectations.
A weak prompt
"Find cases about retaliation" is too broad. The assistant has to guess whether you mean employment, housing, consumer, whistleblower, constitutional, or criminal-law retaliation, plus the jurisdiction, time period, procedural posture, and what kind of answer you need.
Broad prompts are not always bad for brainstorming, but they are risky when you want authority you can cite or rely on.
A stronger structure
- Jurisdiction: where the law must come from.
- Legal issue: the actual doctrine, statute, defense, standard, or procedural question.
- Facts: the few facts that change the legal analysis.
- Output: roadmap, authority list, quote check, source-support note, or next research steps.
- Verification: ask for citations, source text, treatment limits, and uncertainty.
Ask what Descrybe can do
Once Descrybe Legal Engine is connected, one of the best first prompts is simple: ask what Descrybe can do in that chat.
Descrybe gives the assistant a menu of focused legal research tools. The exact tool list can change, so let the assistant tell you what is available now, what each tool is for, and when it would use each one.
As you get more comfortable, you can ask for a specific tool or task. You can also leave tool choice to the assistant: it can break down your question, decide what legal data it needs, call the Descrybe tools that fit the task, and use the returned source material in its answer.
What Descrybe Legal Engine tools are available in this chat? Please explain in plain English what each one does, when you would use it, and when I should ask for that tool specifically instead of letting you choose.What you are building toward
A good prompt is not a magic phrase. It is a short package of context the assistant can use: jurisdiction, issue, posture, safe facts, output, and verification expectations.
In the next lesson, you will turn those ingredients into a roadmap prompt and ask the assistant to plan before it answers.