An operations coordinator at a small nonprofit opens Copilot Chat in VS Code to understand volunteer onboarding docs stored across the repo. The task is read-only research with codebase context, and no files may be edited yet. Which agent selection in Copilot Chat best preserves that read-only research goal?

The correct choice preserves read-only research because Ask mode is designed to answer questions grounded in workspace context, searching across the repo to synthesize volunteer onboarding information without permission to create edits, which directly matches the coordinator's need to understand docs before any change is authorized. The Agent-mode alternative with autonomous steps fails on control despite correct workspace scope, since autonomy invites multi-step tool use including file search, terminal actions and proposed edits that break the no-files-may-be-edited constraint and over-provisions agency for simple comprehension. The Edit-mode alternative for precise file updates fails on intent despite correct broad context, since that mode is optimized to apply scoped rewrites and will steer responses toward diffs and updated content rather than neutral explanation, violating the research-only goal. The Ask-mode alternative limited to currently open files fails on scope despite correct read-only behavior, since restricting grounding to the editor viewport misses dispersed onboarding docs elsewhere in the repository and risks incomplete or hallucinated answers. Exam caveat: distinguish mode permission from prompt wording, because asking a more powerful mode not to edit does not override its underlying capabilities and expected behavior on the test. Practical check before sending: confirm Ask mode is selected, verify workspace or codebase context is attached rather than selection only, and glance that no changed-files preview or edit proposal appears in the response.