agentleFS
KemetraKemetra/Seshat-BI/.claude/skills/powerbi-dashboard-design/SKILL.md

powerbi-dashboard-design

Skill2 starsChanged 4 months agoApache-2.0View on GitHub

The Power BI VISUAL/DESIGN layer of the Tower BI Readiness System. Use when someone asks about dashboard design, page layout, a background/canvas asset, a theme/colors choice, a screenshot critique, a mobile layout, or a Power BI handoff in the Seshat BI repo. This skill ROUTES a design request into exactly ONE of four surfaces (report visuals, external background/canvas, theme JSON, implementation handoff) and opens the matching workflow. It is a router and a vocabulary, not a builder: it never edits a PBIP/PBIR file, never generates DAX, and never invents a metric. The gated "design a dashboard from approved contracts" intent it hands off to the F011/012 dashboard-design verb.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Visuals map to semantic-model fields: Every data-bound visual binds to a field present in the governed semantic model (F010).

    line 85
  2. Distinctive:

    No data-bound dashboard design before the subject area semantic_model_ready is pass (roadmap rule 5).

    line 104

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

powerbi-dashboard-design

The front door of the Power BI design FOUNDATION. The readiness spine defines Stage 6, Dashboard Ready (docs/readiness/dashboard-ready.md): a report is designed AGAINST approved metric contracts, never before them. This skill is the committed design vocabulary that stage reasons WITH -- it classifies any visual-design request into one of four separate surfaces and opens the right workflow, so the four-surface distinction is committed once here instead of being improvised per request.

This skill ROUTES and INSTRUCTS inside Seshat's design layer; it is a nested design router, not a second broad Power BI front door and not an execution-owner selector. Designing a specific dashboard from approved contracts is the F011/012 dashboard-design verb's job. Native PBIR implementation belongs to Microsoft's official powerbi-report-authoring skill when discoverable; F016 is narrower and owns only parked live semantic-model connection/refresh/query/publish execution.

The four surfaces (route to exactly one)

Power BI dashboard design is four separate surfaces. Route every request to exactly ONE; never blend them.

#SurfaceWhat it isAuthoring toolThe rule that keeps it clean
1Report visualscards, charts, slicers, tables/matrices, tooltips, bookmarks, titles, interactions, mobile layoutMicrosoft powerbi-report-authoring or Power BI Desktop after the Seshat gateevery visual binds to a metric contract + a semantic model field; nothing invented
2External background/canvasPNG/SVG/JPG backgrounds, grids, safe zones, static layout containers, exported assetsFigma / Canva / PowerPoint / Illustrator (outside Power BI)background is STATIC STRUCTURE, never data -- no KPI value, no dynamic title baked in
3Theme JSONcolor palette, fonts, visual defaults, page/wallpaper defaults, filter-pane defaults, sentiment colorsa JSON file imported into Power BItheme controls DEFAULTS, never business meaning -- no DAX, no metric, no relationship
4Implementation handoffthe bundle an approved official executor or human uses to build the reportnotes only in this skillthis skill STOPS at the handoff boundary -- no PBIP/PBIR edit

Blending these is the failure mode. Baking a KPI number into a background image mixes surface 1 into surface 2 (a number that never refreshes). Putting a metric definition in theme JSON mixes surface 1 into surface 3 (business logic hidden in a styling file). Editing a PBIP file here crosses into surface 4 (the execution owner's territory). The discipline that keeps the four apart is the value.

Router: request -> workflow

Classify the request, then open the single matching workflow under workflows/.

RequestSurfaceOpen this workflow
New dashboard / page design1workflows/page-blueprint.md
Existing screenshot critique1workflows/screenshot-review.md
Background / canvas request2workflows/background-asset-design.md
Colors / fonts / default formatting3workflows/theme-json-design.md
Propose a formatting plan for an approved page1/3workflows/formatting-plan.md
Chart / card / slicer arrangement1workflows/visual-design-system.md + workflows/page-blueprint.md
Recommend a dashboard pattern for an approved Report Intent1workflows/pattern-recommendation.md
Preview an approved blueprint before any PBIR build1workflows/blueprint-preview.md
Audit a composed report against its committed Report Intent1workflows/dashboard-semantic-audit.md
Final review1workflows/dashboard-qa.md
Mobile-specific layout1workflows/mobile-layout.md
Build / implement in Power BI4workflows/powerbi-handoff.md
Review / verify a built PBIR page4workflows/visual-implementation-review.md

Routing rules:

  • Exactly one surface per request. Name the surface + the intent + the workflow file before doing anything else. Never produce a plan that blends two surfaces.
  • A request that spans two surfaces ("design the page AND its background") splits into two routed sub-tasks (surface 1 + surface 2), each with its own artifact -- never one blended plan.
  • An ambiguous request ("make this look better") is a Principle V STOP: ask which surface is meant rather than inventing a blended plan.

Hard rules (carry into every workflow)

  • Metrics come from metric contracts only. The foundation never invents, defines, or alters a metric -- metric definition is F009's job. A data-bound visual with no backing approved contract is an orphan and MUST NOT be emitted.
  • Visuals map to semantic-model fields. Every data-bound visual binds to a field present in the governed semantic model (F010). A field not in the model is unmapped and is recorded as a blocking reason -- not designed around.
  • Background = structure, not data (surface 2). Never bake a KPI value, a dynamic title, or any other dynamic content into a static background image.
  • Theme = defaults, not meaning (surface 3). The theme controls palette, fonts, visual/page/wallpaper/filter-pane defaults, and sentiment COLORS only; it MUST NOT control DAX, metric definitions, relationships, source mapping, storytelling, or validation. (A sentiment COLOR is the theme's; the sentiment THRESHOLD/RULE is a metric contract, F009.)
  • PBIP/PBIR implementation routes outside this design skill. Native report authoring routes through powerbi-workflows to Microsoft's official skill; bounded Seshat formatting routes to the PBIR adapter; live semantic execution remains parked F016. This skill produces NOTES only.
  • STOP before editing any PBIP/PBIR file. This skill edits no PBIP/PBIR, generates no DAX, changes no SQL, and edits no semantic-model file.

The inherited gate

No data-bound dashboard design before the subject area semantic_model_ready is pass (roadmap rule 5). This feature DEFINES NO new gate and is NOT a second source of truth: the gate, the design-review sign-off, and the dashboard_ready: pass are owned by docs/readiness/dashboard-ready.md and the F011/012 dashboard-design verb. This feature documents and reuses them.

The gate applies to data-bound work (surface 1). Pure-styling work that touches no metric -- a background's safe zones (surface 2), a theme's palette (surface 3) -- is NOT gated and may proceed. Record readiness with the four statuses only (not_started / blocked / warning / pass) plus evidence[] and blocking_reasons[]; never a numeric score, and never self-grant dashboard_ready: pass (that is the verb owner's recorded design-review).

Routing the gated intent: the "design a dashboard from approved contracts" intent routes to the F011/012 dashboard-design verb (it sits O-1 alongside this router). This skill points that intent at the verb; it does not re-implement the gate or design the specific dashboard itself.

See also

  • The prose reference for each surface: docs/powerbi/*.md (visual-design-system, background-assets, theme-json, dashboard-blueprints, visual-qa).
  • The stage this foundation backs: docs/readiness/dashboard-ready.md (the gate to inherit) + docs/readiness/readiness-model.md (the four statuses, no score).
  • The gated verb this router hands off to: the F011/012 dashboard-design skill (.claude/skills/dashboard-design/SKILL.md; spec at specs/012-dashboard-design-skill/).
  • The broad owner selector: public powerbi-workflows plus seshat pbi-mcp doctor.
  • Native report execution owner: Microsoft powerbi-report-authoring when discoverable; F016 remains the parked live semantic connection/refresh/query/ publish adapter.
  • The design-foundation idea lane: docs/roadmap/idea-backlog.md ## Design Foundation -- the categorical cohort where design-layer ideas (strengthens_layer = design-system) land in the idea bank (exploratory, not a roadmap).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.