decision first dashboard

What this skill does Decision-First Dashboard turns KPI-heavy dashboards into interfaces that help people understand what matters, why it matters, and what deserves attention next.

Instead of simply rearranging charts and metric cards, the skill first establishes the decision the dashboard is meant to support, then routes and grounds the available evidence before generating the final interface.

It is designed for dashboards where users currently have to scan many independent metrics before they can understand overall status, investigate an issue, or decide what to do next. Key capabilities Decision-first intake Checks whether a dashboard is actually the right solution and clarifies the recurring decision, action, audience, and review context before rendering begins.

Metric prioritization Routes extracted metrics into roles such as primary signals, diagnostics, exceptions, and drill-down evidence so important information receives appropriate visual prominence.

Semantic preservation Distinguishes trends, distributions, rankings, breakdowns, relationships, and metric clusters so the visual form does not distort what the source data means.

Evidence grounding Keeps visible claims and metrics traceable to source-supported facts and blocks unsupported scores, targets, thresholds, verdicts, or causal claims.

Deterministic output Produces canonical HTML, SVG, and a verification manifest through the skill's compiler rather than allowing the AI agent to redraw the dashboard freely. How it works

1. Confirm the dashboard is worth building The Worthiness Test checks for a recurring decision or monitoring loop, an accountability path, and a meaningful response when an important signal changes.

2. Define the decision A short adaptive intake clarifies what the user needs to determine and what action could change as a result.

3. Ground and route the evidence Source-supported facts are typed, prioritized, and linked to their evidence before they can enter the visual composition.

4. Compile the dashboard The validated state is passed through a deterministic renderer that produces canonical HTML and SVG plus provenance and verification metadata. Why not just prompt an AI to redesign the dashboard? Generic dashboard prompts can produce attractive layouts while silently inventing targets, interpreting direction as performance, dropping important context, or choosing charts based on visual convention rather than data meaning.

Decision-First Dashboard uses explicit intake, semantic typing, grounding, routing, composition, and delivery gates before an output can be considered canonical. Good use cases

• Enterprise dashboards overloaded with KPIs
• Executive or management review dashboards
• Monitoring views where exceptions need clear priority
• Data products that require traceable source evidence
• Existing dashboards that need stronger decision hierarchy rather than simply a visual refresh Current limitations The skill intentionally does not provide free-form AI-generated dashboard layouts. Canonical output must pass the structured Decision-First pipeline, and the renderer uses its own deterministic HTML/SVG visual grammar rather than third-party charting libraries.

Some dense chart-label cases remain an active presentation limitation in v1.1.1.