An independent rerun tests whether the completion claim survives outside the agent's narrative. Use the committed command and inspect the actual result.

Completion claim → Exact command → Fresh result → Accept or reopen
  1. 1Completion claim
  2. 2Exact command
  3. 3Fresh result
  4. 4Accept or reopen

Work through the example

Prompt: List the commands you ran and their outcomes. I will rerun the relevant acceptance check before accepting the change.

One practical example

Use the prompt above on one small, named part of your app. Before accepting an edit, ask the agent to point to the source or configuration that supports its answer. If it cannot locate that context, resolve the missing input before expanding the task.

In the diagram, each arrow represents a handoff you can inspect. Keep the intermediate result rather than jumping directly to the final claim. For example, a reported finding should retain its location, and an executed check should retain its outcome. This gives the next attempt a concrete starting point and makes a misleading completion message easier to challenge.

Apply it to your project

Prompt: List the commands you ran and their outcomes. I will rerun the relevant acceptance check before accepting the change. Then compare the resulting diff with the original request. If it changed a neighboring concern, ask why that change was necessary. Save the result only after the relevant check has run, and record any part that remains unverified.

Evidence and limits

Passing selected tests does not prove untested paths.

This is an educational guide. Its presence in the series does not certify a completed client-specific lab. The series evidence record separates executed checks from exercises and blocked environments.

Inspect the source used in this lesson.

What to do next

Next: Small prompts, one screen at a time, commit between each