auto-agent-router_skill

Overview

This skill automatically routes incoming messages to specialized sub-agents based on command prefixes. It inspects message starts for commands like /coder, /writer, /analyze, and calls sessions_spawn to start the corresponding sub-agent. If no valid command is found at the message start, the main session handles the message.

How this skill works

On every received message the skill strips optional @mentions, then checks whether the message begins with a supported /command. If a matching command is detected it spawns the mapped sub-agent (sessions_spawn) and forwards the task content. If no command is present at the start, routing is skipped and the main session processes the message directly.

When to use it

• You want concurrent, specialized processing by dedicated sub-agents.
• Messages explicitly start with a routing command such as /coder or /writer.
• You need model-specific handling (e.g., a coder model for code tasks).
• You want each task isolated in its own Session for security or concurrency.
• You expect parallel execution by multiple sub-agents.

Best practices

• Require commands to appear at the start of the message for deterministic routing.
• Normalize and remove @mentions before command detection.
• Keep a concise mapping of commands to agents and models in config.json.
• Use /auto only when you accept automatic content-based routing.
• Log spawn events and results to support auditing and debugging.

Example use cases

• User sends: /coder Implement a function — spawns coder agent using a code-specialized model.
• User sends: @bot /writer Draft a blog post — mention is removed, writer agent is started.
• Analyst sends: /analyze Summarize dataset — analyst agent runs with analysis model.
• Team triggers: /devops Deploy service — devops agent executes operational steps.
• Fallback: simple question without command is handled by the main session without spawning a sub-agent.