What problem does it solve? Coordinating multiple AI agent profiles on shared work is error-prone: orchestrators invent nonexistent assignees, workers lose context between retries, and handoffs lack structure. This Skill provides the decomposition playbook for orchestrator profiles and the lifecycle, handoff, and retry guidance for worker profiles in a Hermes Kanban multi-agent setup. ## Core Features & Use Cases - Orchestrator decomposition playbook: Discover available profiles with hermes profile list, sketch task graphs, fan out independent lanes in parallel, and gate dependent cards with parents=[...] links. - Worker lifecycle guidance: Orient with kanban_show, send meaningful heartbeats, produce structured kanban_complete summaries with metadata, and write actionable kanban_block reasons. - Retry and recovery diagnostics: Interpret prior run outcomes (timed_out, crashed, spawn_failed, reclaimed) and apply reclaim, reassign, or model-change recovery strategies. - Use Case: A user asks to evaluate a database migration. The orchestrator creates two parallel research cards (cost, performance) assigned to a research profile, plus a synthesis card with both as parents assigned to an analyst profile, then reports the task graph back to the user. ## Quick Start Ask the orchestrator to list available profiles with hermes profile list, then decompose your request into Kanban cards assigned to those profiles.