coordinatorlisted

@markdownai v1.0 ## When to use coordinator mode Use when a task requires parallel or sequential sub-agent work: analysis across multiple files, multi-step pipelines, or any work that benefits from isolation. ## Boot 1. Run /fleet-dashboard to confirm all agents are reachable and Postgres is up. 2. Identify which agents (heimdallr, hanuman, shiva, etc.) should own each subtask. ## Fan-out pattern 1. Create a fork for the initiative: fork_create(app_id: heimdallr, title: "initiative name", created_by: heimdallr) 2. Dispatch subtasks in parallel: agent_task_submit(app_id: heimdallr, task: "...", agent: hanuman) agent_task_submit(app_id: heimdallr, task: "...", agent: shiva) 3. Poll completion: agent_task_status(app_id: heimdallr, task_id: ...) 4. Synthesize results into a KB atom: kb_ingest(app_id: heimdallr, title: "...", summary: "combined findings") 5. Close the fork: fork_merge(app_id: heimdallr, fork_id: FORK-XXXX, outcome_note: "...") ## Priority groups (for routing decisions) | Group | Agents | Use for | |----------|-------------------------------------------|--------------------------------| | OPERATOR | willow, ada, steve | Infrastructure, system tasks | | ENGINEER | heimdallr, hanuman, opus, kart, shiva, ganesha | Feature work, analysis | | WORKER | Others | Narrow, bounded subtasks | ## Rules - Always chec