This set builds vocabulary for describing autonomous, multi-step coding agent workflows in Copilot.
0 / 5 completed
1 / 5
At standup, a dev mentions switching Copilot from suggesting individual lines to autonomously completing a multi-step coding task across several files. Which mode fits?
Agent mode lets Copilot autonomously plan and execute a multi-step task, editing multiple files and running commands, rather than only suggesting the next few tokens as a developer types. This shifts Copilot from reactive autocomplete toward a more independent, goal-directed workflow. It still typically surfaces changes for review before finalizing.
2 / 5
During a design review, the team wants to review each file Copilot's agent mode plans to touch before it proceeds. Which capability supports this?
Agent mode typically surfaces a plan or diff preview of intended changes before finalizing them, letting a developer catch scope issues or unintended file touches early. This upfront visibility mirrors the trust-building pattern used by other agentic coding tools. Skipping this review increases the risk of unreviewed changes landing.
3 / 5
In a code review, a dev notices agent mode ran a terminal command to install a dependency as part of completing the task. Which capability does this represent?
Agent mode can invoke tool or command execution, such as installing a dependency or running a build step, as part of autonomously completing a task, going beyond generating suggested text. This execution capability is what distinguishes an agent from a passive suggestion engine. It also raises the stakes for reviewing what commands were actually run.
4 / 5
An incident report shows agent mode modified a file outside the scope described in the original task prompt. What practice reduces this risk?
A narrowly scoped prompt combined with reviewing the complete diff before accepting changes catches out-of-scope edits before they land. Vague or overly broad task descriptions increase the chance the agent touches more than intended. This discipline applies across agentic coding tools generally, not just Copilot's agent mode.
5 / 5
During a PR review, a teammate asks how agent mode differs from Copilot's standard inline suggestions. What is the distinction?
Standard inline suggestions predict the next tokens as a developer types in one file, while agent mode plans and carries out a broader task spanning multiple files and potentially executing commands. This agentic scope shifts the developer's role toward directing and reviewing rather than typing every change manually. It is the same shift seen across other agent-mode features in modern coding tools.