Dual-Channel Worker Polling

Problem

In an AI-agent worker loop, a single inbox poll usually covers only explicitly delivered private tasks. Messages that should also trigger action—@worker mentions, discussions, and broadcasts in chat—arrive through a separate surface. Running two independent pollers tends to introduce duplicate consumption, unstable ordering, scattered priority logic, and missed messages when one channel overwrites the other. worker-poll unifies that entry point.

How It Works

The skill folds both channels into one polling and processing path:
- Ch1 Inbox: private task channel; checks pending tasks first;
- Ch2 Chat: group-chat channel; checks @my mentions, broadcasts, and discussion context;
- Merge and deduplicate: collapses messages from both channels that point to the same task or intent into one execution item;
- Priority sort: orders execution by task priority instead of raw arrival time, reducing the chance that lower-priority chatter interrupts critical work.

It relies on Dashboard API endpoints such as GET inbox and GET chat, with authentication via Authorization: Bearer {TOKEN}. This is useful as a replacement for single-inbox polling in worker main loops, especially for agents that must monitor both group chat and inbox. The boundary is also clear: it focuses on polling aggregation, deduplication, and priority ordering, not semantic task resolution, tool execution, or conflict handling. If both channels issue conflicting instructions for the same task, an upper-layer agent policy should still make the final decision.