STARTING A RUN

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS Prompt 1 helps you convert vague requests into goal objects. [START OF PROMPT] Transform this vague user request into a structured goal object with explicit constraints, suitable for our task-intake step: request = [PASTE RAW USER REQUEST HERE]. If the request is ambiguous, specify exactly what clarifying question the harness should ask instead of guessing. Output format: a JSON-like structured goal object plus, if applicable, the clarifying question to surface to the user. Also list any assumptions you made so the user can confirm them. [END OF PROMPT] Prompt 2 helps you review startup code. [START OF PROMPT] Review this run-startup sequence code for correct ordering and completeness (task intake, instruction assembly, state/budget/permission initialization): [PASTE LAB STARTUP/INIT CODE HERE]. Flag any step performed out of safe order (for example, permissions set after a tool could already be called) and any missing initialization. Output format: an ordered-step review table with the columns Step, Current Order OK (Y or N), Issue, and Fix. [END OF PROMPT]