Practice

Write the complete prompt for a model that receives technical incident details and must draft a client-ready update. The update needs to balance transparency, calm tone, and actionable next steps while avoiding internal jargon leakage.

You are writing the full prompt for a model that receives messy vendor invoice text. The model must extract the important fields into a strict JSON object. The hidden evaluator cares about schema compliance, normalized values, and whether missing fields are represented consistently rather than hallucinated.

Write the full prompt for a model that receives messy meeting notes and must turn them into a concise recap plus a structured list of action items. The output should separate the high-level summary from the individual tasks so another teammate can scan it quickly.

Design a prompt that reads an incoming support complaint and outputs a compact JSON triage decision. The response should identify urgency, routing, and whether the customer needs a follow-up without adding extra prose.

Create a prompt that answers a customer question using only the policy excerpts included in the input. The model should give a helpful answer while citing the exact policy sections it relied on.

Design a prompt that receives a customer support incident timeline and makes the model analyze the likely root cause without guessing. The response should separate evidence, reasoning, and conclusion clearly: include quoted evidence from the timeline, a reasoning section that explains how the evidence supports the conclusion, and a final answer with the most likely root cause, confidence level, and next action.

Write a prompt for a model that receives account-call notes and turns them into an executive-ready summary. The response should separate the customer goals, delivery risks, and follow-up commitments instead of mixing everything together.

Design a prompt that makes the model pause and ask clarifying questions when a request is underspecified. The answer should explain what is missing, list the questions, and avoid overcommitting before enough information is available.

Create a prompt that rewrites a rough internal message into something calm, direct, and professional without losing the operational facts. The output should show the rewritten message and the issues it corrected.

Write the full prompt for a model that compares two status updates and extracts where they agree, where they conflict, and what remains uncertain. The response should be machine-readable so another system can act on it.

Create a prompt that turns the model into a lightweight planning agent. The model will receive a user request plus a small tool catalog and should decide what tool call to make before returning a final answer. This mirrors interview questions where structure, explicit planning, and clean tool arguments matter more than writing code.

Design a prompt that synthesizes multiple short documents into a concise executive brief without losing provenance. The output should summarize the situation, call out key points, and surface open questions while citing the source IDs used.

Create a prompt that turns an analytics request and table schema into a careful SQL planning response. The model should explain the approach, write a SQL draft, and call out assumptions instead of pretending the schema is perfectly clear.

Write the prompt for a model that judges an answer against a rubric and returns a structured scoring result. The output should be machine-readable, balanced, and useful for coaching.