What problem does it solve? When a user request spans multiple specialists, an orchestrator agent can be tempted to do the work itself or create poorly structured tasks. This Skill provides a decomposition playbook that enforces routing work through a Kanban board instead of executing it directly, preventing silent failures from unknown assignees and broken dependency chains. ## Core Features & Use Cases - Profile Discovery Before Planning: Requires checking which worker profiles actually exist on the machine (via hermes profile list or asking the user) before assigning cards, since the dispatcher silently drops unknown assignees. - Task Graph Decomposition: Guides splitting multi-lane requests into parallel independent cards and gated dependent cards using parents=[...] links in kanban_create calls. - Goal-Mode Cards: Supports goal_mode=True for open-ended cards, wrapping workers in a judge-evaluated loop until acceptance criteria are met or the turn budget is exhausted. - Stuck Worker Recovery: Documents reclaim, reassign, and model-change workflows for tasks flagged with warnings or hallucination events. - Use Case: A user asks to "research Postgres migration costs and performance, then draft a decision memo." The orchestrator creates two parallel research cards, a synthesis card gated on both, and a memo-drafting card gated on the synthesis, each assigned to real profiles on the setup. ## Quick Start Ask the orchestrator to break down a multi-part goal into Kanban cards assigned to your available worker profiles and report the resulting task graph.
