Diagnostics

A diagnostic defines the context for one or more investigation sessions. It specifies the AI's behavior (via a system prompt), which tools the AI has access to, and who can view the diagnostic and its results.

Creating a Diagnostic

From the Diagnostics list, click Create Diagnostic. Provide:

• Name — A descriptive name for the investigation context.
• Domain — Optional category for the diagnostic (e.g., "Cost Savings", "Quality Metrics"). You can select an existing domain from other diagnostics or type a new one. Domains help organize diagnostics across the team.
• Entity — Optional entity to scope the investigation. When set, all sessions in the diagnostic automatically receive a filter instruction telling the AI to restrict queries to this entity's data.
• Vintage — Optional free-form text identifying the data period (e.g., "2024 Q3", "CY2025").
• Description — Optional context about the diagnostic's purpose.
• System Prompt — Instructions that define the AI's role and focus area.
• Tools — Which capabilities the AI has access to during sessions.
• Editors — Architect-level (or higher) users or user groups granted full edit access to the diagnostic and all its content, with the same rights as the owner.
• Viewers — Users or user groups granted read-only access to the diagnostic and all its content.

Diagnostic List

The diagnostic list shows all diagnostics you own, can edit, or can view. Each diagnostic displays its name, description, and the last active date — the most recent time any session in the diagnostic was updated. This helps you quickly identify which diagnostics have recent activity.

System Prompts

The system prompt is the primary way to direct the AI's behavior. It appears at the beginning of every conversation and should describe:

• The AI's role and expertise area.
• The domain or dataset being investigated.
• Any specific instructions about how to approach the analysis.

The full prompt sent to the AI is assembled from multiple parts. The Prompt Preview on the diagnostic form shows the complete structure:

1. Your system prompt
2. Keyword instructions (how the AI signals tool use)
3. Tool-specific prompts (instructions for each enabled tool)
4. Upstream dependency context (findings and dispositions from upstream sessions, if the session has dependencies — see Upstream Context Injection)
5. Disposition schema context (if the session defines one)
6. Entity filter context (if the diagnostic is associated with an entity — instructs the AI to filter queries to the entity's data)
7. The user's starting request

Tools

See Tools Reference for the full list.

Sharing

By default, a diagnostic is private — only the owner can see it. To share a diagnostic, add individual users or user groups as editors or viewers on the diagnostic form.

• Editors have the same rights as the owner: they can edit the diagnostic, create and manage sessions, run AI conversations, edit findings, and manage who else has access. Only Architect-level users (or higher) can be added as individual editors; any user group can be added as an editor group. When a user group is added as an editor, all members of that group receive editor-level access.
• Viewers have read-only access to the diagnostic, including all sessions, conversation logs, and findings. Any authenticated user or user group can be added as a viewer. When a user group is added as a viewer, all members of that group receive read-only access.

Removing all editors and viewers (both individual and group) makes the diagnostic private again.

For diagnostics created by a scan, access is also granted through scan-level sharing. A user who is an editor or viewer of the scan gains the corresponding access to all of the scan's diagnostics, in addition to any sharing configured directly on the diagnostic. See Opportunity Scan — Sharing for details.