The grounded prompt skeleton
Rung 1 of 10 · RTO — Role · Task · Output
1 · Learn the move · RTO — Role · Task · Output
The one structure under every defensible deal prompt: Role, Context, Task, Constraints, Output, Examples — with grounding baked into Constraints every time: work only from the facts provided, trace every number to an input, mark gaps UNKNOWN. Six brackets to fill. Everything else here is this skeleton, practiced to reflex.
ROLE: You are a [specific CRE expert]. CONTEXT: [The asset, the audience, the document.] TASK: [The one concrete thing to produce.] CONSTRAINTS: [Length, tone, AND the grounding rules: only the facts I pasted, every figure traced to an input or labeled an assumption, gaps marked UNKNOWN — never filled.] OUTPUT: [Abstract / memo / email / analysis.] EXAMPLES: [One grounded sample, if you have one.]
2 · Your turn — you write the prompt
Take a deal task you handed AI this week without guardrails — a lease summary, an investor email, a quick property blurb. Rewrite the request using all six parts, with the grounding rules built into the constraints.
Remember: the AI sees only your prompt — not this page. If the situation isn't in your prompt, it doesn't exist.
Optional — these shape the output when you run your prompt below, not your score.