Prerequisites
- Scope: Admin privileges required.
- Environment: Desktop/browser UI (websocket sessions) only. Not supported via API or programmatic runs.
- Model Recommendation: Use high-capability LLMs. Smaller models may fail to invoke the tool.
Enable the Skill
- Navigate to Settings → Agent Skills in the admin sidebar.
- Click the slider icon in the top right corner to open the agent skill settings modal.
- Toggle on Allow agent to ask clarifying questions.
- Set Max questions per turn (Range:
1–10, Default:3).
Technical Specifications
Input Types Reference
| Type | Format Options / Parameters |
|---|---|
| Text input | Free-form text, URL, number, date, email |
| Single-select choice | Dropdown/list (supports optional “Other” write-in field) |
| Multi-select choice | Checkboxes/list (supports optional “Other” write-in field) |
Runtime Behavior
- Execution Flow: Agent pauses mid-turn -> Renders interactive card (paginated if >1 question) -> User responds/skips -> Agent resumes with answers injected into context.
- State Persistence: Survey transcripts (Q&A) are appended to chat history and injected into the prompt history for subsequent turns. Read-only survey cards are rendered upon reloading the conversation.