Give follow-up feedback that changes the right thing

Specific feedback identifies the file, section and desired correction while protecting what is already right. It is more useful than a general request to make the output better.

Start with a small handoff

Select a role by its working guide and deliverables, then verify the connected AI tool and workspace. Keep the first run small enough that you can inspect every output. The desktop needs to stay running while work executes. Marketplace sign-in, model access and file-generation tools are separate setup requirements; a success in one does not prove the others are ready.

Inputs for this workflow: The generated artifact and a short list of observed issues.

Work through the task

1. Separate factual corrections from style preferences.
2. Specify what should remain unchanged.

A brief you can adapt

In summary.md, correct the period to the supplied date range and separate estimates from actuals. Keep the approved structure and save a revised copy.

The filenames above are examples. Replace them with your actual inputs and destination before submitting the task.

Review the deliverable

Compare the revised file with the original and check both requested changes and accidental regressions.

Where this approach can fail

Session continuation depends on the chosen tool. Include essential source and filename context even in a follow-up.

For the related desktop setup, see first task.