A repository team wants Copilot cloud agent to help remediate failing integration tests. The team does not want the agent to start code changes until a human has reviewed the proposed plan. The repository already uses issues, pull requests, branch protection, and required checks. Which four actions should you perform in order? 1. Ask Copilot to produce a structured remediation plan 2. Assign the issue to Copilot immediately 3. Review the proposed plan against repository constraints 4. Start the agent task only after approval is recorded 5. Merge the resulting draft pull request 6. Record approval after human review

The correct sequence prioritizes planning and human oversight before execution. First, "Ask Copilot to produce a structured remediation plan" (1) leverages the agent's capabilities to outline a solution. Next, "Review the proposed plan against repository constraints" (3) ensures the plan aligns with team policies and technical requirements, which is crucial given the team's desire for human review before changes. Then, "Record approval after human review" (6) formally documents the team's acceptance of the plan. Finally, "Start the agent task only after approval is recorded" (4) initiates the agent's work, adhering to the requirement that code changes begin only after human approval. Option 2 → 3 → 6 → 4 is incorrect because assigning the issue to Copilot immediately (2) before a plan is even generated or reviewed bypasses the critical planning and review steps. Option 1 → 6 → 2 → 5 is incorrect because recording approval (6) before reviewing the plan (3) is illogical, and merging a draft pull request (5) without prior approval of the plan and execution is premature. Option 3 → 1 → 4 → 6 is incorrect because reviewing the plan (3) before Copilot has even produced one (1) is impossible, and starting the agent task (4) before approval (6) violates the team's constraint.