Automations
Tool set: automationsDefault: Disabled
Create deliberately scoped event-to-action rules. Arbitrary triggers, outbound sends, delays, task updates, and agent-to-agent triggers are excluded from the public surface.
Tools
| Tool | Description |
|---|---|
| create_automation | Create an event → action rule |
| list_automations | List all automation rules |
| delete_automation | Delete an automation |
Trigger Events
| Event | When |
|---|---|
| task.created | A task is created |
| task.updated | A task is updated |
| task.deleted | A task is deleted |
| note.created | A note is created |
| note.updated | A note is updated |
| document.ready | A document finishes processing |
| asset.visibility_changed | An image is published or made private |
| asset.deleted | An image is deleted |
Actions
| Action | Description |
|---|---|
| create_task | Create a task |
| create_note | Create a note |
| store_memory | Store a memory |
| run_prompt | Agent processes a prompt |
Template Variables
Use {{event.title}}, {{event.status}}, etc. in action configs to reference the triggering event's data.
Task events cannot create tasks, and note events cannot create notes. This blocks direct self-loops; causal chains also stop after depth 3.
Reliability and compatibility
Each event/rule pair has a stable execution key. Prompt actions pause and resume around the agent job. Approval-required outbound tools inside a prompt run are blocked instead of creating an approval request with no safe resume path. Legacy rules outside this matrix remain visible for review but cannot be re-enabled unchanged.
Limits
| Limit | Value |
|---|---|
| Automations per agent | 30 |
| Actions per automation | 10 |
| Maximum causal depth | 3 |