A developer is mid-task in Copilot Agent Mode, which has been editing several files autonomously, when they decide the remaining work needs careful edit-by-edit approval instead of autonomous action. They want to continue refining the same change conversationally but regain per-edit control. What is the most appropriate way to proceed?
- AStay in "Agent Mode" but add a custom instruction telling Copilot to pause and request confirmation before it writes to any file during the rest of the session.
- BSwitch to "Plan Mode" so Copilot drafts an ordered breakdown of the steps, then have it apply every step at once without a further review stage.
- COpen the "GitHub Copilot CLI" in a terminal and re-describe the change there so each shell command is shown for approval before it runs against the files.
- DSwitch to "Edit Mode" in the same Copilot Chat panel and describe the remaining changes, reviewing each proposed edit before it is applied to the files. Correct
Why A is wrong: Agent Mode is built to act autonomously and run tasks to completion, and a custom instruction does not convert it into a reliable per-edit approval surface, so it does not meet the need.
Why B is wrong: Plan Mode produces an ordered breakdown before coding rather than applying reviewable edits, so it is tempting for structure but does not give the edit-by-edit approval being sought.
Why C is wrong: The CLI suggests shell commands for terminal tasks, not editor file edits with per-edit review, so it is the wrong surface for refining an in-editor code change.
Why D is correct: Edit Mode applies multi-file changes from a natural-language instruction but surfaces each edit for the developer to accept or reject, which gives the per-edit control they now want.