Give your agent a task.
Know when it’s done.
Practical guides for ChatGPT dot and Meta Muse. Define the responsibility, keep consequential decisions in your hands, and verify the outcome in the original system.
Independent resource · Not affiliated with OpenAI or Meta
Different agents.
The same finish line.
Start with the product you use, then choose a responsibility with clear access boundaries, approval points and evidence.
What needs doing?
Pick a situation you recognize. These Meta Muse playbooks give the agent a clear job—and give you a clear way to check it.
Use Muse to review a higher internet bill
Your promotional rate ended or your monthly bill increased.
Use Muse to find and cancel an unused subscription
A recurring charge appears for a service you no longer use.
Use Muse to follow up on a missing refund
A seller says it issued a refund, but you cannot find the credit.
Use Muse to check a post-purchase price drop
The price fell soon after you bought an item.
Use Muse to review a roofing quote
A contractor sent a roof repair or replacement estimate.
Use Muse to review an HVAC replacement quote
You received a quote to replace or install heating or cooling equipment.
A prompt is only
the beginning.
Other sites show what an agent might do. We focus on the part that matters after you press send: boundaries, approvals, and evidence.
See how we write and review guidesUse the verification checklist
Prepare
Know the trigger, materials and limitations before you start.
Delegate, with limits
Copy a task instruction that tells the agent when to stop and ask.
Verify
Check the original account, record or recipient—not just the agent's reply.
Give ongoing work a boundary.
Start with the ChatGPT dot responsibility brief, or use the platform-neutral checklist before approving an agent's result.
Build a daily work brief →