The Commander is a meta-agent that operates above the sprint level. It is dispatched when a user creates an initiative — a large goal that spans multiple sprints (e.g., “rebuild the authentication system” or “launch the marketplace feature”). The Commander breaks the initiative into a sequenced sprint plan, monitors progress across sprints, and adjusts the plan as sprints complete or fail.
Responsibilities
Initiative decomposition — break a large goal into a sequenced set of sprints
Dependency ordering — determine which sprints must complete before others can start
Progress tracking — monitor sprint outcomes and update the initiative plan accordingly
Re-planning — when a sprint fails or produces unexpected output, revise the remaining sprint plan
Initiative completion — declare the initiative complete when all sprints finish and all acceptance criteria are met
Skills
Skill
Description
initiative.plan
Decomposes a large goal into ordered sprints
sprint.sequence
Manages the queue of pending sprints for an initiative
progress.track
Monitors sprint states and updates the initiative timeline
plan.revise
Revises the remaining sprint sequence based on completed sprint outcomes
initiative.complete
Validates all acceptance criteria and closes the initiative
When dispatched
INTAKE: when a sprint is flagged by the Captain as epic complexity, the Commander is invoked to decompose it into multiple sprints
Initiatives vs. sprints
Concept
Scope
State machine
Sprint
Single feature or task
INTAKE → THINK → PLAN → BUILD → SHIP → COMPLETE
Initiative
Multi-feature goal
PLANNING → IN_PROGRESS → COMPLETE (with N sprints inside)
Sample system prompt excerpt
You are the Commander agent for Defiant 2.0.
A user has defined a large initiative that cannot be completed in a single sprint.
Your job is to decompose it into a sequenced set of sprints.
Initiative goal:
<goal>{{ initiativeGoal }}</goal>
Project context:
<context>{{ projectContext }}</context>
Rules for decomposition:
1. Each sprint must be independently shippable -- it produces value on its own.
2. No sprint may take longer than 2 hours (estimated). If a unit of work would
exceed this, split it further.
3. Order sprints so each depends only on sprints that precede it.
4. The first sprint should establish the foundational infrastructure others build on.
5. Include a "stabilization" sprint at the end to address any issues that surface
during the initiative.
Output: an ordered list of sprint goals, each with acceptance criteria and
