An architect is designing an agent handoff protocol from a Claude orchestrator to a specialized code review subagent. What information must the handoff packet include to enable the subagent to operate effectively without requiring further orchestrator interaction?
Select an answer to reveal the explanation.
Short Explanation
You need a handoff packet that gives the subagent the task objective, relevant context, success criteria, and output format. That lets it review code without asking you or the orchestrator for more info. Just sending the raw diff leaves it blind to standards and goals.
Full Explanation
An effective agent handoff packet must include four elements: task objective — what specifically the code review should accomplish; relevant context — PR description, coding standards, related files, or constraints; success criteria — what constitutes a complete review (e.g., security check, style compliance, logic correctness); and output format — how results should be structured for the orchestrator to consume programmatically. The raw-code-diff-only option is insufficient — without coding standards and objectives, the subagent cannot calibrate the review appropriately. The full-orchestrator-history option passes irrelevant context, wastes tokens, and may expose sensitive information from earlier in the orchestrator's session. The orchestrator-system-prompt-copy option is incorrect — the subagent has its own specialized system prompt and should not replicate the orchestrator's behavioral constraints.