AI Skills¶
AI Skills are user-defined instructions that are automatically provided to the AI Assistant as additional context. Use them to supply domain-specific knowledge, business rules, database schemas, or any other information that helps the assistant work more effectively with your data.
Skill Scopes¶
Every skill has a scope that determines when it is included:
| Scope | Belongs To | Included In |
|---|---|---|
| Global | Workspace | Every conversation in the workspace |
| Report | A specific report | Conversations opened from that report |
| External Database | A specific external database | Conversations for any report that uses a data source connected to that database |
Skills are resolved automatically when you open the AI Assistant -- global skills are always included, report and database skills are included when relevant. You do not need to activate them manually.
Managing Skills¶
Global skills are managed from the sidebar under the AI section.
Report skills are managed from the Report tab in the Report Editor toolbar via the Skills button. You can link an existing global skill to the report or create a new skill scoped exclusively to that report.
Database skills are managed from Workspace Configuration > Connections > External Databases via the Skills button on each database. The same link/create options apply.
When removing a skill: if it was created locally (report-scoped or database-scoped), removing it deletes the definition. If it is a linked global skill, removing it only unlinks it -- the global definition remains available elsewhere.
The assistant automatically loads database skills when constructing queries for reports that use the corresponding database.
Writing Effective Skills¶
Skill content is plain text or Markdown -- write it as you would write instructions for a colleague.
- Be specific and concise. Focused content produces more reliable results than long, general instructions.
- Use structured formats. Tables, lists, and headings help the assistant parse the information.
- Include examples. Show the output format or query pattern you expect.
- Avoid contradictions across global, report, and database skills.
Example: Database Schema Skill¶
## Customer Database Schema
### customers
- id (int, PK)
- name (varchar)
- email (varchar)
- created_at (timestamp)
- region_id (int, FK → regions.id)
### orders
- id (int, PK)
- customer_id (int, FK → customers.id)
- total_amount (decimal)
- order_date (date)
- status (varchar: 'pending', 'shipped', 'delivered', 'cancelled')
### regions
- id (int, PK)
- name (varchar)
- country_code (varchar)
Always JOIN through region_id when filtering by region name.
Use order_date (not created_at) for date-range filters on orders.
Example: Report Layout Skill¶
This report follows the company brand guidelines:
- Page header: company logo (top-left), report title (centered), date (top-right)
- Section headings: 16px, bold, color #1a365d
- Body text: 11px, color #2d3748
- Tables: alternate row shading with #f7fafc, header row #2b6cb0 with white text
- Charts: use the "corporate" color palette from the theme
Access Control¶
| Permission | Level | Allows |
|---|---|---|
| AI | Read | Use the AI Assistant (read-only chat) |
| AI | Full Access | Full use of the AI Assistant |
| Workspace Configuration | Full Access | Create, edit, and delete skill definitions and assignments |
Notes¶
- Skill changes take effect in the next conversation or when the chat is reconnected.
- Very large skills consume model context window capacity and may reduce response quality. Keep them as concise as possible.
- Skills are workspace-specific and cannot be shared across workspaces.