What problem does it solve? Dispatched Kanban workers often fail silently or produce unusable handoffs: they claim cards they never created, block with vague reasons like "stuck", call interactive tools in headless runs, or repeat the exact approach that already timed out. This Skill provides the operational detail that prevents those failures. ## Core Features & Use Cases - Structured handoff patterns: Concrete kanban_complete summary and metadata shapes for coding, research, and review tasks, plus a review-required flow using kanban_comment plus kanban_block so human reviewers get full context. - Workspace and tenant handling: Rules for scratch, dir:, and worktree workspace kinds, and tenant-prefixed memory entries to prevent cross-tenant context leaks. - Retry and edge-case diagnostics: How to interpret prior run outcomes (timed_out, crashed, spawn_failed, reclaimed, blocked) and avoid repeating failed paths, plus heartbeat and block-reason conventions. - Use Case: A worker retrying a crashed task inspects kanban_show, sees outcome: "crashed", reduces memory footprint, completes with structured metadata, and claims only the card IDs captured from actual kanban_create return values. ## Quick Start Load the kanban-worker skill and follow its guidance to complete your dispatched Kanban task with a structured summary, verified created_cards, and a clear block reason if human input is needed.
