FreeYour four-agent team is readyClaim Your AI Team
Docs

Build

Reviewing agent work

Check results and boundaries through the conversation before relying on an agent for important work.

3 min read

Use a controlled first task

Your own agent conversations live in Agents. Use a clearly scoped first request and review the outcome there and in Activity.

Make the first request useful

  1. 01

    State the outcome

    Describe a concrete result and any constraints.

  2. 02

    Keep the first action small

    For consequential work, ask for a limited first step or a proposal before execution.

  3. 03

    Answer questions

    Supply missing information and review any connection or permission request.

  4. 04

    Review the evidence

    Check the delivered work, its sources, and the recorded status.

  5. 05

    Give feedback

    Tell the agent what to change and whether it is a job-specific correction or an ongoing preference.

Check more than the happy path

  • Ask what the agent should do when required information is unavailable.
  • Confirm it respects a restriction or approval requirement.
  • Verify important facts against the source.
  • Treat a failed or blocked job as unresolved until its status and outcome are clear.