Shaduf.Research preview
AI Web Design & Frontend Skills Catalog/For a dashboard, define the data task before adding a skill

6 October 2026 · Task-fit source check

For a dashboard, define the data task before adding a skill.

A dashboard-app builder asked for a better frontend skill set after disappointing results. That is one request, not proof that another skill will fix the screen. The useful first decision is what the dashboard is failing to show.

Choose one route

SituationStart withCheck
Existing components and tokens already govern the screenA project brief naming the user's decision, data fields, source of truth, units, freshness, filters, and required states. No added skill by default.Does the rendered screen reuse the system and make status and comparison clear?
The screen needs a full dashboard design workflowPracticalSwan Frontend Design 2.0, pinned at afbb609 as a conditional candidate. Its author calls for scanning, comparison, status visibility, units, freshness, and empty or error states.Give it the actual project context; inspect narrow and desktop widths, keyboard operation, and missing-data states.
The product's visual identity is genuinely openAnthropic Frontend Design, pinned at 41bbe19 for visual direction. Its instructions emphasize palette, typography, and layout.Do not let a new palette override an existing system or let visual novelty obscure the data task.
The implementation exists and named files need reviewVercel Web Design Guidelines, pinned at 063bee94 for a file-and-line review.It fetches mutable rules at runtime. A clean file review is not a rendered-result or accessibility verdict.

A fair first trial

Give the same agent and model the same dashboard project, brief, data, and acceptance checks. Compare a no-skill run with one candidate skill, not a stack. Inspect data clarity, component and token reuse, empty and error states, keyboard behavior, viewport changes, and cost. We have not run that comparison.

Evidence limit: These routes are editorial inferences from pinned instructions and one user question. They do not rank outcomes or show adoption. Our sixteen listings still have no pool-run no-skill comparison.

Search published pools, pages, reports, and evidence.