Playbook Generator Agent Reference
The Playbook Generator Agent runs in Playbook Building Mode on playbook and component editor pages. Starting in Turbine 26.2.0, Playbook Builder v2 adds clarifying questions and expanded supported scenarios for building on the canvas.
Starting in Turbine 26.3.0, the agent can create and modify multiple flows in one request. Select flows on the canvas or name them in your prompt, then review proposed changes with Accept all / Reject all. You can still work on one flow at a time by selecting it on the canvas or naming it in your prompt.
For procedures, see Create and Modify Playbooks with Hero AICreate and Modify Playbooks with Hero AI. For clarifying questions, progress detail, and multi-flow review, see the same topic.
Quick Reference β Operation Matrix
Entity | Create | Modify | Delete | Enable | Disable | Reorder | Move (Cross-Flow) |
|---|---|---|---|---|---|---|---|
Playbook metadata (name, description) | β | Yes | β | β | β | β | β |
Flow (with at least one trigger or action) | Yes | Yes | Yes | Yes | Yes | β | No |
Schedule (cron) trigger | Yes | Yes | Yes | β | β | β | β |
Flow event trigger | Yes | Yes | Yes | β | β | β | β |
Webhook trigger | Yes | Yes | Yes | β | β | β | β |
Record event trigger | No | Filter only on existing | No | No | No | β | β |
Playbook button trigger | No | Filter only on existing | No | No | No | β | β |
Trigger schema (webhook or flow event) | Yes | Yes | With trigger | β | β | β | β |
Action | Yes | Yes | Yes | Yes | Yes | Yes | No |
Connection (on-success, on-failure, on-complete) | β | Yes | Yes | β | β | β | No |
Component (Building Mode) | β | Yes | β | β | β | β | β |
Playbook Metadata
The agent can read and update playbook name and description. It can also read created by, modified by, and timestamp metadata when you ask who created or last changed the playbook.
Operation | Supported |
|---|---|
Read playbook name, description, authorship, timestamps | Yes |
Update playbook name or description | Yes |
Enable or disable the entire playbook at product level | No β use flow enable or disable instead |
Save, duplicate, copy, export, or package the playbook | No β use the UI |
Example prompts: Who created this playbook? Β· When was this playbook last modified? Β· Update the playbook description to β¦
Flows
Multi-flow editing (26.3.0): The agent can create or modify multiple flows in one request, or work on one flow at a time when you select a flow on the canvas or name it in your prompt.
Multiple flows appear as separate graphs on the same canvas. The agent resolves which flow you mean from your wording, the selected flow on the canvas, or from multi-flow context across the playbook.
Scenario | Supported (26.3.0) | Notes |
|---|---|---|
Create a new flow with at least one trigger or action | Yes | Empty flows are not supported |
Create multiple new flows in one request | Yes | ο»Ώ |
Edit one existing flow | Yes | Default when you name a flow or have one selected |
Edit multiple flows in one request | Yes | Select on canvas and/or name flows in the prompt |
Delete a flow by name, position, or content | Yes | Removes the flow graph |
Enable or disable a flow by name or selection | Yes | Sets flow enabled state |
Enable or disable all flows | Yes | Prompt with βall flowsβ; agent updates each flow |
Rename a flow to a unique title | Yes | UI flow name = flow title |
Rename a flow to a title already used by another flow | No | Agent asks before editing |
Copy or move a flow to another playbook | No | ο»Ώ |
Connect an action in one flow to an action in another | No | Agent refuses cross-flow edges |
Explain what a flow does | Yes | Read-only; loads all flows |
Bulk Accept all / Reject all of pending AI changes | Yes (UI) | Canvas summary footer when changes are pending |
Selected flow / selected nodes: Canvas selection scopes edits to the selected flow unless you name other flows. For action-level disable/enable/delete/modify, the agent prefers selected nodes (or explicitly named actions). Prefer flow titles over position language (βfirstβ, βleftmostβ).
Flow disable vs action disable: Disabling a flow pauses the entire graph. Disabling an action skips only that step while the rest of the flow can still run.
Triggers
Five trigger types exist. The agent can create, modify, and delete schedule, flow event, and webhook triggers. Record event and playbook button triggers can only be filtered on flows where they already existβthe agent cannot create, replace, or delete those trigger types.
Trigger | Create | Modify Metadata | Modify Filter (if) | Delete | Enable / Disable Trigger |
|---|---|---|---|---|---|
Cron / Schedule | Yes | Yes | β | Yes | β |
Flow Event | Yes | Yes | Yes | Yes | Yes |
Webhook | Yes | Yes | Yes | Yes | Yes |
Record Event | No | No | Yes (existing only) | No | No |
Playbook Button | No | No | Yes (existing only) | No | No |
Schedule Trigger
Scenario | Supported |
|---|---|
Create a flow with a schedule (for example, every 7 minutes or daily at 6 AM) | Yes |
Change the cron expression on an existing schedule | Yes |
Replace a schedule with a webhook or flow event trigger | Yes |
Add a schedule to a flow that had another trigger type | Yes |
Example prompts: Ingest data every 7 minutes Β· Create a flow that runs daily at 6 AM and searches for records with status New
Flow Event Trigger
Scenario | Supported |
|---|---|
Create a new flow with a new flow event trigger | Yes |
Create a flow reusing an existing flow event definition from another playbook | No |
Create a flow with a trigger similar to an existing flow event | Yes |
Rename flow event trigger id or name | Yes |
Add or modify trigger filter conditions (if) | Yes |
Map actions to $event.data from the trigger schema | Yes |
Delete a flow event trigger | Yes |
Replace a flow event trigger with cron or webhook | Yes |
Webhook Trigger
Scenario | Supported |
|---|---|
Create a new flow with a webhook trigger | Yes |
Create a webhook flow similar to an existing webhook flow | Yes |
Rename a webhook trigger | Yes |
Add or modify webhook filter conditions (if) | Yes |
Map automation to $event.data.body fields | Yes |
Replace webhook with schedule or flow event | Yes |
Delete webhook trigger only (keep rest of flow) | Yes |
Configure webhook authentication (shared secret, format, value) | No |
Generate webhook URL | No β use the UI |
Record Event and Playbook Button Triggers
Scenario | Supported |
|---|---|
Create a new record event or playbook button trigger | No |
Replace an existing trigger with record event or playbook button | No |
Modify filter conditions on an existing record or button trigger | Yes |
Replace record event or playbook button with cron, webhook, or flow event | Yes |
Trigger Conditions (All Conditional Triggers)
The agent can build and modify filter conditions including:
- Numeric comparisons (for example, $gte, $lte)
- String and regex match ($match)
- Logical AND / OR across fields
- Record field change detection (previous vs record)
- Array and nested field conditions
Trigger Schemas
Webhook and flow event triggers use a trigger schema that documents $event.data field shapes. Schedule triggers do not use a per-flow trigger schema.
Scenario | Supported |
|---|---|
Read trigger schema before mapping $event.data | Yes |
Create or update trigger schema when adding webhook or flow event fields | Yes |
Create or replace schema for record or playbook button triggers | No |
Example prompts: Map $event.data.alert_name to Case Name on webhook ingest Β· Create a webhook flow with transformation of webhook body fields
Actions
Action Lifecycle
Scenario | Supported | Notes |
|---|---|---|
Add a new action (add, insert, another, new, one more) | Yes | Creates a new node; does not change existing actions |
Modify an existing named action | Yes | Only the targeted action |
Delete an action by name, description, or selection | Yes | Rewires connections |
Enable or disable an action | Yes | Uses action disabled state |
Rename an action key and title in the flow | Yes | Updates references in connections |
Edit action inputs (limits, filters, mappings) | Yes | ο»Ώ |
Edit a Python script body (specific lines) | Yes | ο»Ώ |
Insert an action between two connected actions | Yes | Rewires on-success paths |
Reorder two actions in the execution chain | Yes | Updates entrypoints and connections |
Move or copy an action to another flow | No | ο»Ώ |
Group actions into a sub-component | No | UI only |
Mark an action input as sensitive | No | UI only |
Configure connector retry, delay, or limit | No | ο»Ώ |
Add vs Modify
Your wording | Agent behavior |
|---|---|
add, insert, another, additional, new, one more | Add a new action; do not edit existing |
change, update, edit, set β¦ on [named action] | Modify only that action |
Ambiguous and the same action type already exists | Agent asks a clarifying question before editing |
all flows, every instance, both flows | Not in 26.2.0 β edit one flow at a time |
Native Actions
The agent can add, configure, and delete these native action types:
- Condition, loop, parallel
- Create or update variable
- Search records, create record, delete records
- Scripts, transformations, HTTP
- Hero AI native action
These native actions are not fully supported through prompts (configure them in the UI if needed):
- Export records
- Emit event
- Swimlane Intelligence
- Send notification
- CSV get rows
Components and Connectors
- Components (preferred): Add, configure, and remove tenant components as sub-actions in a playbook flow.
- Connector actions: Add, configure, and remove connector actions; add or change the asset used by a connector action.
- Prefer components unless you request a connector directly or no component exists for the task.
Component Building Mode (component editor):
- Modify user-defined input and output schemas
- Publish fields to properties in the output schema
- Cannot add or modify existing interfaces (intents)
Connections
Each action has at most one incoming connection and either (one on-success + one on-failure) or (one on-complete).
Scenario | Supported |
|---|---|
Change connection type (on-success, on-failure, on-complete) | Yes |
Connect an action in one flow to an action in another flow | No |
Multiple outgoing on-success from one action | No |
Multiple incoming connections to one action | No |
Mix on-success or on-failure with on-complete on the same action | No |
Query and Explain (Read-Only)
Scenario | Supported |
|---|---|
What does this playbook or flow do? | Yes |
Which flows use Hero AI, search records, or a specific application? | Yes |
Who created or modified the playbook? | Yes |
Explain action logic in plain language | Yes |
Read live application records (outside automation context) | No |
Read playbook run history or logs | No |
Test or run the playbook or individual actions | No |
Clarification and Ambiguity
Starting in Turbine 26.2.0, the agent uses clarifying questions when your prompt is ambiguousβfor example, when multiple applications, components, or fields match, when add vs modify is unclear, or when the target flow is unspecified in a multi-flow playbook. See Create and Modify Playbooks with Hero AIClarifying Questions in Create and Modify Playbooks with Hero AI.
Scenario | Behavior |
|---|---|
Multiple resources match your intent | Agent asks before implementing |
Add vs modify not clear | Agent asks |
Target flow ambiguous | Agent asks or uses selected flow |
You explicitly said add | Agent adds; does not ask add vs modify |
Bulk change (all flows, every instance) | Not in 26.2.0 β one flow at a time |
Content Locking and Concurrent Sessions
Locked by Another User
When a playbook is locked for editing by another user, Hero AI can still describe and query the playbook but cannot apply changes. If a lock occurs while generation is running, the UI may show suggested changes with Reject or Copy changes onlyβno Accept. Refresh the page after the lock is released to resume editing.
Same Playbook in Multiple Tabs or Windows
Turbine does not prevent the same user from opening the same playbook in multiple browser tabs or windows. Each session can edit and save independently; the last save wins, and changes from other sessions can be overwritten. This is separate from multi-user content locking above.
The same behavior applies when another product workflow opens the playbook builder in a new window (for example, Create a Playbook or Recreate Playbook on an AI SOC case record). Do not click those actions multiple times for the same playbook, and finish one builder session before you start another.
For manual editing on the canvas, see How to Create a PlaybookHow to Create a Playbook. For Playbooks OverviewPlaybooks Overview, see the Turbine User Guide.
Validation and Safety
After the agent changes a flow, the platform validates the playbook graph. Fix validation errors in the canvas or with follow-up prompts before you save.
Not Supported
Scenario | Why |
|---|---|
Move or copy an action between flows | Each flow is isolated |
Rename a flow to a duplicate display title | Name collision |
Create, replace, or delete record event or playbook button triggers | Guardrail β filter-only on existing |
Webhook authorization or shared secret configuration | Out of scope β use the UI |
Enable or disable the entire playbook (product-level toggle) | Use per-flow enable or disable |
Export, duplicate, save, or package the playbook | UI operation |
Run or test the playbook | Use Test Console |
Create or modify applications, fields, or field values | Out of scope |
Mark inputs sensitive or configure field masking | UI security settings |
Group actions into a sub-component | UI operation |
Cross-flow action wiring | Platform isolation |
Use variables from assets unless mapped to connector actions | Asset mapping rule |
Reference playbook or component runs in prompts | Out of scope |
Answer general Turbine documentation questions while in Building Mode | Use Hero AI Companion outside the canvas |
Notes and Considerations
- Prompts in Hero AI native actions, Python scripts, and JSONata transformations are not validated and may be incorrect.
- Complex (multi-level, multi-group) filters in loops, conditions, search records, and trigger conditions might not be fully correct.
- Filters by attributes with predefined list values are not supported in Search Records.
- Create Variables must not be used inside loops.
- When modifying existing actions, the agent changes only what you name or select on the selected flow.
- For end-to-end validation after AI changes, use the Test ConsoleTest Console.
See Also
- Create and Modify Playbooks with Hero AICreate and Modify Playbooks with Hero AI β procedures, clarifying questions, progress detail
- Hero AIHero AI β overview of Hero AI features and building modes
- Playbooks OverviewPlaybooks Overview β canvas layout, flows, and data model
- Hero AI CompanionHero AI Companion β general companion outside the canvas