Skip to content
AssureSwarm Docs

Agent Operating Guide

/coach-dashboard-create

Build a configurable dashboard, one underlying query feeding stat, table, chart, and filter widgets, submitted as a suggestion for review.

Create a dashboard that visualizes tenant data. A dashboard has one underlying GraphQL query that feeds the whole view, and each widget renders a slice of that result: stat rows, tables, charts, and filter bars. The skill gathers the name, visibility, and query, then walks you through the widgets one at a time, and submits one suggestion via suggest_change for a human to approve: the dashboard isn’t created until it is.

When to use

Section titled “When to use”

When you want to create, build, design, author, or set up a dashboard, report view, or data-visualization page. For a one-off answer you don’t need to keep, use /coach-query-data instead. This skill creates a dashboard; editing an existing one is a separate update flow.

Inputs

Section titled “Inputs”
Flag Required Notes
--name <text> No Dashboard title. Asked for if absent.
--description <text> No One-paragraph purpose.
--visibility <private|public> No Defaults to private; asked if absent.
--query <graphql> No A pre-built GraphQL string for the whole dashboard.
--widgets <json> No A pre-built widgets array. Gathered interactively if absent.

Example

Section titled “Example”

/coach-dashboard-create --name "Audit Status" asks for the purpose and visibility (defaulting private), builds the query behind the view and tests it with query_data before baking it in, then walks you through widgets, say a stat row of counts, a status filter bar, and a table of audits, and submits a create suggestion you approve in Canvas.

Good to know

Section titled “Good to know”
  • The query is tested before it’s baked in. Building widgets on an unverified query is how you get empty widgets, so the skill runs the query via query_data first.
  • Widgets read the result by dot-path, and any table or chart that filters must reference a filter bar’s id: a mismatch surfaces as a validation error.
  • Visibility defaults to private (creator plus admins). Public puts the dashboard in every user’s list, so ask for it explicitly.
  • Chart type follows the data: bar for categories, line for time series, pie or donut for proportions (five slices or fewer), gauge for “X of N”, horizontal bar for long labels.
  • This is create-only. Editing an existing dashboard is a separate flow; there’s no native drill-down, though a table row can link out.
Section titled “Related”