Skip to main content
Good prompts tell Doe the goal, sources, output, constraints, and validation expectations. This matters most for larger requests where agents may use files, integrations, research, spreadsheets, connected data, and code sandboxes.

The Pattern

Vague vs Specific


Ask For End-to-End Work

Doe is strongest when you ask for a finished outcome, not just a short answer.

Use @mentions

Type @ to reference the exact context Doe should use.

Ask Doe to Search First

If you do not know which file, thread, or app has the answer, say that. Doe can search before it reads deeply.

Ask for a Durable Output

Doe can create or edit artifacts, not just answer.

Use Validation Language

Tell Doe how careful it should be.

Make Work Repeatable

Add schedule language when you want a Loop.

Redirect Anytime

You can send a message while agent work runs to narrow scope, change format, add context, or stop before an external action.

Task Playbooks

Turn prompts into repeatable work

Preview Pane

Watch agents work alongside chat

Search

Learn the difference between chat search, Library search, and content search

Sharing

Share responses, conversations, files, and folders

Loops

Turn prompts into recurring tasks