Intelligent Task Dispatcher

Problem to Solve

In multi-agent or multi-worker setups, routing a task cannot rely only on text matching. Workers may be offline, out of quota, underperforming, or limited by specific capability boundaries. If the orchestrating agent decides ad hoc each time, it can misroute work to unavailable workers, ignore shared team context, create inconsistent batch dispatches, or assign critical work to poorly matched nodes.

How It Works

cnbll-task-dispatcher is a task-routing skill for agent workflows. It triggers when a task needs to be sent to a specific worker, when routing by worker specialty is required, or when distributing tasks to multiple workers in bulk. It reads config/worker-roster.json in v2 format, including org_context and dispatch_rules, and matches workers from dynamic rules instead of hardcoded mappings.

Before dispatch, it injects context such as who the worker is, the team objective, and the worker’s historical output, so the worker can understand the broader task background. It also filters by status and performance: offline or quota-exhausted workers are skipped, and low-performing workers are used only when no other worker is available. The task is then sent through Dashboard API: POST inbox.

Boundaries and Notes

This skill fits agent orchestration setups that already have a worker roster and an inbox API. It is not a good fit when there is no worker-roster.json, no clear worker capability model, or no explicit status, quota, or performance signals. Dispatch quality depends heavily on roster maintenance and rule clarity.