maestro-mission

Maestro Mission

You are orienting at or working inside a maestro mission. Missions are the planned-execution layer: milestones, features, assertions, checkpoints, handoff launches.

The task system is separate. Use maestro-task for the daily queue.

---

When to activate

• "Show missions", "what missions are in flight", "what am I working on"
• "What feature am I on", "list features for mission X"
• "Report feature progress", "mark feature X as passed/blocked"
• "Read the prompt for feature X"
• Agent needs to render the mission-control TUI for the user

Top-level orientation

maestro status --json

Prints session id, active missions, current task counts, and mission-control summary. Always a safe first call.

Mission layer

maestro mission list --json
maestro mission show <missionId>

Mission lifecycle transitions (draft -> approved -> executing -> completed), milestones, checkpoints, and assertion semantics live in ./reference/mission-lifecycle.md.

Feature layer

maestro feature list --mission <id> --json
maestro feature prompt <featureId> --mission <id>
maestro feature update <featureId> --mission <id> \
  --status <pending|assigned|in-progress|review|done|blocked> \
  --report @report.json

feature prompt returns the briefing a feature-scoped agent reads before execution, with memory context auto-injected. Use it to understand what a feature expects, or to re-render after memory changes.

The --status value tracks the feature's own state (pending, assigned, in-progress, review, done, blocked). The attached --report is where per-assertion results live (passed / failed / blocked / waived). The two are distinct axes: a feature can be done with a mix of passed and waived assertion results. See ./reference/assertions.md for the report shape.

Memory corrections

Capture a correction rule for future sessions:

maestro memory-correct "use bun not npm" --trigger "package,install,npm"

--trigger is a comma-separated list of keywords. When a future agent's work matches a trigger, the correction is injected as context.

Mission-control TUI

maestro mission-control is a read-only TUI. It has an agent-friendly preview mode that prints plain-text frames to stdout.

maestro mission-control --preview --size 120x40 --format plain
maestro mission-control --preview all --size 120x40 --format plain
maestro mission-control --render-check --size 120x40
maestro mission-control --json --size 120x40

Preview modes:

• --preview: renders the current default screen as plain text.
• --preview [screen]: renders a specific screen (dashboard, features, dependencies, config, memory, graph, agents, dispatch, events, tasks, timeline, principles, help). Use all to render every screen sequentially.
• --render-check: validates every preview screen renders without errors, returns a JSON report.
• --format plain|ansi: plain-text for agent-consumable frames or ANSI-styled output (default auto-detects TTY). --format is ONLY valid with --preview.
• --json: print a structured snapshot instead of a rendered frame. Use --json alone; it is mutually exclusive with --preview.

Read-only contract. Preview, JSON, and render-check paths must not mutate state. Use for inspection only. Full flag reference: ./reference/mission-control-tui.md.

Reference

• ./reference/mission-lifecycle.md: draft/approved/executing transitions, checkpoint semantics
• ./reference/mission-control-tui.md: flag reference, --screen options, format matrix
• ./reference/assertions.md: assertion state machine and update rules