Coordinator Tools Reference
The coordinator session has 30 built-in tools it uses to orchestrate workers, manage tasks and teams, author artifacts, and control loops. These are NOT slash commands — they are LLM-callable functions the coordinator invokes autonomously. Source: src/coordinator/toolHandlers.ts → domain builders in src/coordinator/tools/. Use describe_feature for the live schema. Authoring tools save artifacts; call reload_registry after save_workflow or manual edits, and activate_crew separately.
Generated for agents-fleet v0.60.0 · 30 coordinator tools.
Worker Management
| Tool | Inputs | Description |
|---|---|---|
spawn_worker | name (string; required) — Unique name for the worker agentagent_type ("general-purpose" | "explorer" | "coder" | "reviewer" | "tester" | "custom" | "coordinator" | "researcher" | "workflow-runner"; required) — Type of agent to spawnprompt (string; required) — The task/prompt for the worker to executetask_id (string; optional) — Task ID (e.g., "T-1") this worker is assigned to.model (string; optional) — Model to use for this worker (defaults to session model)cwd (string; optional) — Working directory override (default: auto-created worktree for coders)role (string; optional) — Explicit role name from the active crew (e.g., "fleet-coder").skills (string[]; optional) — Explicit atomic skill names to compose with the role.skill (string; optional) — @deprecated Use role instead.permissionOverrides (object; optional) — Per-spawn permission overrides for this worker.requiredCapabilities (("shell" | "write" | "meta")[]; optional) — Capabilities the assignment requires.localArtifacts (string[]; optional) — Local files or directories already supplied for the worker to inspect.completionContract (object; optional) — Structural evidence required before idle may count as success.mergeOnSuccess (boolean; optional) — Issue #419: when false, the worker's worktree commit is preserved on its branch (no auto-merge into local master, worktree not removed) so a downstream workflow stage (e.g. /github-issue-worker-autopilot handoff → gitBridge) can pick up the changes and open a PR. | Spawn a worker for a specific task. |
send_message | to (string; required) — Agent ID to send message to (e.g., "coder@my-team") or "*" for broadcastmessage (string; required) — The message content to sendsummary (string; required) — 5-10 word summary of the message | Send a message to a running worker agent or use "*" to broadcast to all team members. |
stop_worker | agent_id (string; required) — ID of the agent to stop | Stop a running worker agent |
list_workers | status ("all" | "active" | "completed" | "failed"; optional) — Filter workers by status (default: all) | List all worker agents and their current status |
Task Management
| Tool | Inputs | Description |
|---|---|---|
task_create | subject (string; required) — Short title for the taskdescription (string; required) — Detailed description of the taskowner (string; optional) — Agent ID who owns this taskblocked_by (string[]; optional) — Task IDs that must complete before this task can startdod_ids (string[]; optional) — Definition-of-Done item IDs (e.g. ["DoD-1","DoD-3"]) this task is responsible for satisfying. | Create a new task with subject, description, and optional dependencies |
task_update | task_id (string; required) — ID of the task to update (e.g., "T-1")status ("pending" | "in_progress" | "completed" | "blocked" | "deleted"; optional) — New status for the tasksubject (string; optional) — Updated subjectowner (string; optional) — Agent ID to assign as ownerresult (string; optional) — Result text when completing a taskadd_blocks (string[]; optional) — Task IDs that this task should blockadd_blocked_by (string[]; optional) — Task IDs that should block this taskdod_ids (string[]; optional) — Definition-of-Done item IDs (e.g. ["DoD-1","DoD-3"]) this task satisfies. | Update a task status, owner, result, or dependencies |
task_list | status ("all" | "pending" | "in_progress" | "completed" | "blocked"; optional) — Filter by status (default: all)owner (string; optional) — Filter by owner agent ID | List all tasks with their status and dependency information |
task_get | task_id (string; required) — ID of the task to retrieve | Get detailed information about a specific task |
task_stop | task_id (string; required) — ID of the task to stop | Stop a running task |
get_ready_tasks | — | Get tasks that are ready to start (all dependencies completed) |
Team Management
| Tool | Inputs | Description |
|---|---|---|
team_create | team_name (string; required) — Name for the new teamdescription (string; required) — Description of the team purpose | Create a named team for organized multi-agent collaboration |
team_delete | team_name (string; required) — Name of the team to delete | Delete a team. |
team_add_member | team_name (string; required) — Name of the team to add member tomember_name (string; required) — Name for the new team memberagent_type ("general-purpose" | "explorer" | "coder" | "reviewer" | "tester" | "researcher"; required) — Type of agent for this memberprompt (string; required) — Initial task/prompt for the membermodel (string; optional) — Model for this member (defaults to session model) | Add a specialized agent as a member of a team |
team_list_members | team_name (string; optional) — Team name (defaults to active team) | List all members of a team with their status |
broadcast | message (string; required) — Message to broadcast to all team memberssummary (string; required) — 5-10 word summary of the broadcast | Send a message to all team members |
Memory & Definition of Done
| Tool | Inputs | Description |
|---|---|---|
dod_update | dod_id (string; required) — DoD item ID (e.g., "DoD-1" or just "1")passed (boolean; required) — true if the DoD item passed verification, false if failedevidence (string; required) — Evidence or reason for pass/fail | Record a Definition of Done item as passed or failed. |
fleet_remember | note (string; required) — A clear, concise learning or pattern to remember (e.g., "Workers cannot call slash commands — use tools instead")source (string; optional) — Where this learning came from (e.g., worker name, file path, or observation) | Persist a learning, pattern, or important observation to project memory (.fleet/context/memory.md). |
Authoring
| Tool | Inputs | Description |
|---|---|---|
save_role | name (string; required) — Kebab-case role name (e.g. "fleet-coder").description (string; required) — Short human-readable description of the role.agentType ("general-purpose" | "explorer" | "coder" | "reviewer" | "tester" | "custom" | "coordinator" | "researcher"; required) — Agent type that determines worker permissions.systemPrompt (string; required) — The full system prompt body (Markdown) for this role.model (string; optional) — Preferred model for this role (optional).whenToUse (string; optional) — Guidance on when the coordinator should use this role.outputFile (string; optional) — Optional output filename produced by this role. | Persist a v2 Role artifact to the project tier (.fleet/roles/<name>.role.md). |
save_skill | name (string; required) — Kebab-case skill name (e.g. "vitest-patterns").description (string; required) — Short human-readable description of what the skill covers.systemPrompt (string; required) — The skill body — knowledge/instructions composed into a role at spawn time.tags (string[]; optional) — Optional tags for categorization (e.g. ["testing", "typescript"]).allowOverride (boolean; optional) — If true, user/project tier may override a same-named bundled skill. | Persist a reusable atomic skill (knowledge pack) to the project tier under .fleet/skills/. |
save_crew | name (string; required) — Crew name in kebab-case (e.g. "feature-tdd-crew").description (string; required) — One-line human description of what the crew does and when to use it.topology ("silent" | "hub" | "mesh"; required) — Communication topology between agents.workflow (string; required) — Workflow name driving the crew (e.g. "freeform", "tdd", "feature-skills-crew-tdd").coordinator (object; required) — The coordinator that drives this crew.members (object[]; required) — Crew agents — at least one required.synthesis (object; optional) — Optional synthesis agent that runs after the others to produce a final summary.mcpServers (string[]; optional) — Crew-wide MCP servers available to all agents unless overridden per-agent.tier ("user" | "project"; optional) — Where to save the crew (default: "project"). | Persist a v2 Crew (workflow + coordinator + agents) to disk as a *.crew.md file under the project (or user) tier. |
save_workflow | name (string; required) — Workflow name (e.g. "feature-tdd").content (string; required) — The FULL .workflow.md markdown you authored — YAML frontmatter (name, description, stages, …) plus body.tier ("user" | "project"; optional) — Where to save the workflow (default: "project").description (string; optional) — Optional human description (informational; the authoritative description lives in the content frontmatter). | Persist an authored .workflow.md (full markdown content) to disk under the project (or user) tier, after validating it with the workflow linter. |
activate_crew | name (string; required) — Kebab-case crew name to activate (must match /^[a-z][a-z0-9-]*$/). | Activate a saved crew by name. |
reload_registry | — | Re-read all workflows, crews, roles, skills, and gates from disk (bundled + user + project tiers). |
Discovery
| Tool | Inputs | Description |
|---|---|---|
list_features | kind ("workflow" | "skill" | "role" | "crew" | "command" | "tool"; optional) — Optional filter — only list features of this kind. | List the fleet capabilities registered in this session: workflows, roles, atomic skills, crews, built-in slash commands, and coordinator tools. |
describe_feature | name (string; required) — The feature name to describe (e.g. "code-review", "fleet-coder").kind ("workflow" | "skill" | "role" | "crew" | "command" | "tool"; optional) — Optional kind hint to disambiguate when a name exists across kinds. | Describe a single fleet feature by name (workflow, role, atomic skill, crew, built-in command, or coordinator tool). |
Loop & Workflow Control
| Tool | Inputs | Description |
|---|---|---|
signal_loop_complete | reason (string; required) — Why the loop is complete (1-500 chars) | Signal that the active loop should stop. |
signal_workflow_abort | reason (string; required) — Why the workflow should be aborted (1-500 chars)runId (string; optional) — Optional workflow runId (e.g. wf-1a2b3c4d). | Abort a workflow (e.g. /feature or /github-issue-worker-autopilot). |
Channels
| Tool | Inputs | Description |
|---|---|---|
send_attachment | path (string; required) — Absolute or cwd-relative path to the file to send.caption (string; optional) — Optional caption text rendered alongside the attachment.kind ("auto" | "document" | "photo" | "audio" | "video" | "voice"; optional) — Attachment kind. | Send a file attachment (photo, audio, video, voice, or document) to all paired remote chats. |
Other
| Tool | Inputs | Description |
|---|---|---|
dod_generate | items (object[]; required) — The DoD items to author, in order.plan_slug (string; optional) — Optional plan slug to write into.mode ("append" | "replace"; optional) — append (default): keep existing DoD items and continue numbering after them. | Author the Definition of Done for the active plan. |
dod_coverage | plan_slug (string; optional) — Optional plan slug to inspect. | Report DoD coverage for the active plan: which tasks lack a DoD mapping (dod_ids) and which DoD items have no task, then apply the soft execute gate (AGENTS_FLEET_DOD_GATE: off|warn|enforce, default warn). |