Skip to main content
Good Strawberry prompts are practical. Start by telling the companion what you want to do, then ask it to plan the work with you before it starts executing.

​
Start with the outcome

For important tasks, begin with plain language:
This works well because Strawberry can help define the scope, choose useful sources, spot missing context, and turn a vague request into a concrete plan.

​
Example starter prompt

After the companion asks questions, you can agree on details such as:
  • Which sources to use
  • How large the first sample should be
  • What the final output should look like
  • What requires approval
  • How the companion should verify the work

​
Give the companion permission boundaries

Say what it can do automatically and what it must ask about. Useful boundaries:
  • “Read only. Do not send, update, delete, or buy anything.”
  • “Draft emails but do not send them.”
  • “Create a sample of 10 rows first.”
  • “Ask before opening paid tools.”
  • “Use my CRM only for deduplication.”

​
Ask for verification

When accuracy matters, ask for:
  • Source links
  • Confidence notes
  • Rows with missing data
  • Assumptions
  • A summary of how the companion checked its work

​
Use memory and skills

Put stable facts in memory. Put repeatable process in skills. Then prompts can be shorter. If a chat handles a workflow well, ask:
If the workflow should happen on a schedule, ask:
Strawberry can turn the successful pattern into reusable instructions or a routine, then you can review and adjust it.

​
Common mistakes

  • Asking for “everything” without a scope
  • Omitting the source of truth
  • Not saying what format you want
  • Letting a companion infer approval rules
  • Asking for a huge batch before testing a sample
  • Repeating a successful workflow manually instead of turning it into a skill or routine

​
Better follow-ups

Instead of restarting, steer the current task: