Problem

Webhook mappings (hooks.mappings) support action: "wake" | "agent", but there is no way to specify the target agent. All webhook-triggered agent runs always go to the default agent.

This makes it impossible to use webhooks in multi-agent setups — e.g., routing incoming emails to a dedicated "email-agent" with its own tools, workspace, and persona.

Proposed Solution

Add an optional agentId field to hooks.mappings[*]:

{
  hooks: {
    mappings: [
      {
        match: { path: "/email" },
        action: "agent",
        agentId: "email-handler",  // <-- NEW
        messageTemplate: "New email from {{from}}: {{subject}}",
        deliver: true,
        channel: "telegram",
        to: "123456789"
      }
    ]
  }
}

Semantics

• agentId optional (default: default agent)
• Must exist in agents.list[].id
• Agent run uses workspace, tools, and model config of the target agent
• Session key is scoped under the target agent (agent:<agentId>:hook:...)

Use Cases

1. Email processing: Dedicated agent with restricted tools (no shell access, only email reply capabilities)
2. GitHub webhooks: CI/CD agent with repo access, isolated from personal assistant
3. Support inbox: Separate agent per webhook source with different personas/system prompts

Alternative (less elegant)

If agentId in mappings is too complex: A separate endpoint /hooks/agent/:agentId that takes the agent from the URL path.