Tasks

The task system lets agents delegate work to sub-agents that run in separate context windows. Each sub-agent does its work — reads files, explores code, runs searches — then returns a compact result. Your main agent's context stays clean.

Two tools. No orchestration graphs, no DAGs, no state machines.

Tools

delegate

Send work to a sub-agent. It runs in its own context window and returns when done.

{
  "name": "delegate",
  "parameters": {
    "agent": "researcher",
    "message": "Find the top 5 Rust HTTP frameworks by GitHub stars"
  }
}

The target agent must be in the calling agent's members list. See delegation scope below.

collect

Gather results from delegated work.

{
  "name": "collect",
  "parameters": {
    "agent": "researcher"
  }
}

Returns the sub-agent's compact result.

Delegation scope

Agents can only delegate to agents listed in their members config:

---
members: ["researcher", "coder"]
---

An empty members list means the agent cannot delegate — delegation tools won't be available. When members is set, the agent catalog is automatically provided and the delegation tools are added to the agent's tool whitelist.

Skills vs agents

Skills inject behavior into the same context. Agents isolate work into separate contexts. Two distinct tools for two distinct problems.

Use skills when you want to change how your agent behaves. Use delegation when you want to offload work without polluting your agent's context.

What's next

• Agents — configuring delegation with members
• Skills — prompt-level behavior injection
• Event loop — how task events flow through the daemon