maestro:mission-planning

Mission planning transforms a raw idea into a persisted plan by decomposing the goal into 3–7 sprint-sized milestones, assigning features to agents, and defining verification steps. It outputs the exact handoff command to launch the first external agent without executing it.

Agent orchestration coordinates task execution across codex-cli, claude-code, subagent, or human operators. Mission planning uses a decision table to match each feature to the correct agent type, including a different instance rule for review milestones to ensure proper verification.

Mission planning captures explicit scope boundary constraints with reasons, creating do not touch rules for downstream agents. This prevents scope drift by persisting these limitations directly into the mission file before the handoff command is generated.

Yes, mission planning supports agent-type matching for codex-cli, claude-code, subagent, and human operators. It applies a different instance rule for review milestones, ensuring the exact maestro handoff command targets the correct launchable execution step.

Defining verification steps requires breaking a confirmed goal into milestones with 1–5 outcome features each. Mission planning structures these features with explicit dependencies and concrete verification steps, persisting the complete mission plan to disk via maestro mission create.

No, mission planning drafts the exact maestro handoff command for the first launchable codex or claude execution step without executing it. The operator must interactively confirm intent and manually run the outputted command to start the external agent.