Vibe Coding Skills¶
BrickflowUI works especially well with modern AI coding tools when the assistant is given a clear product shape and a few framework-specific guardrails.
This page collects the repo-local skill files we ship for that purpose.
Start here first¶
The canonical high-context guide for AI tools is:
SKILL.mdat the repo root
That file is the authoritative reference for:
- the runtime mental model
- state and event handling
- component usage
- page composition
- loading, theming, and deployment expectations
The smaller skill files under skills/ should be treated as focused overlays,
not replacements for the root guide.
Why this matters¶
Generic "build me a dashboard" prompts often produce one of two bad outcomes:
- notebook-style pages with weak product structure
- generic frontend advice that ignores BrickflowUI's Python-first runtime
The skills in skills/ keep the assistant closer to the framework's real strengths: data and AI portals, internal tools, branded shells, smoother controls, and Databricks-aware delivery.
Included skills¶
| Skill | Path | Best for |
|---|---|---|
| BrickflowUI App Starter | skills/brickflowui-app-starter/SKILL.md |
Starting new apps from the right example and keeping structure clean |
| BrickflowUI Data + AI Portal | skills/brickflowui-data-ai-portal/SKILL.md |
Dashboards, copilots, pipeline command centers, and analytics workspaces |
| BrickflowUI Geometric UI | skills/brickflowui-geometric-ui/SKILL.md |
Glassmorphism, geometric product layouts, visual showcases, and image-led premium surfaces |
| BrickflowUI Polish + QA | skills/brickflowui-polish-and-qa/SKILL.md |
Responsiveness, loading, spacing, interaction quality, and release polish |
| BrickflowUI Databricks Delivery | skills/brickflowui-databricks-delivery/SKILL.md |
Packaging, deployment, managed-platform handoff, and enterprise delivery review |
Recommended workflow¶
- Read the repo-root
SKILL.md. - Pick the closest real example from
examples/. - Start with
skills/brickflowui-app-starter/SKILL.md. - Move to the domain skill that fits the app shape.
- If the app is highly design-led, add
skills/brickflowui-geometric-ui/SKILL.md. - Finish with
skills/brickflowui-polish-and-qa/SKILL.md. - Use
skills/brickflowui-databricks-delivery/SKILL.mdbefore shipping or demoing to a platform team.
End-user priority¶
When you ask an AI tool to work inside BrickflowUI, it should optimize for these people first:
- data platform teams
- AI application teams
- analytics engineers
- internal tools teams
- enterprise evaluators deciding whether the framework can support a real rollout
That priority matters because it changes the kind of app the assistant should build:
- less widget collage
- more operator clarity
- stronger shell and workflow structure
- honest loading and governance behavior
- more realistic examples and demos
What a strong BrickflowUI agent should know¶
A strong BrickflowUI-oriented AI assistant should understand:
- the root
SKILL.mdis the main source of truth - the app shell comes before the chart set
- controls must change visible state
- images and media are legitimate first-class surfaces, not decoration
- loading and empty states are part of the product contract
- examples are part of the framework quality bar
- dark and light modes both matter
- geometry-heavy or editorial designs should still remain maintainable and state-safe
BrickflowUI-specific design rules¶
For dashboard and portal work¶
- Start with
Hero,SectionHeader,Sidebar,TopNav,Card,Grid, andTable. - Use
StatusStrip,GaugeChart,Timeline, andPipelineGraphwhen the page needs operator understanding, not just KPI display. - Keep the fold readable in under five seconds.
For visual showcase or geometry-heavy work¶
- Prefer image-led composition over trying to fake premium art with plain boxes.
- Use rounded shells, pill actions, and calm gradients deliberately.
- Document any missing primitive honestly instead of hiding it in brittle CSS hacks.
- If the design is driven by one hero image, make it large enough to dominate the fold.
For AI/copilot work¶
- Keep chat and support telemetry close together.
- Use drawers, timelines, and tables as explanation layers.
- Avoid making the chat pane the only source of truth if structured data is available.
Prompt patterns that produce stronger results¶
Product-grade portal¶
Use the BrickflowUI app starter and data + AI portal skills. Start from the closest repo example, keep the shell product-grade, add real loading states, and make every filter visibly change a chart, table, or workflow surface.
Premium design-led surface¶
Use the BrickflowUI geometric UI skill. Build a premium glassmorphism showcase in pure Python, prefer local SVG assets for hero art, keep the interactions real, and document any framework primitives that should become first-class for pixel-perfect fidelity.
Release-quality pass¶
Use the BrickflowUI polish + QA skill. Audit this app like an enterprise evaluator: responsiveness, dark/light parity, empty/loading states, control smoothness, and example honesty all matter more than adding new components.
Example prompts¶
Build a data platform portal¶
Use skills/brickflowui-data-ai-portal/SKILL.md and adapt examples/data_pipeline_command_center/app.py into a multi-page portal for data and AI teams. Keep the shell product-grade, add KPI cards, a pipeline graph, cost view, and an operations table.
Start from the best baseline¶
Use skills/brickflowui-app-starter/SKILL.md and tell me which existing example is the best baseline for a secure internal AI operations workspace. Then implement the first working page.
Polish before demo day¶
Use skills/brickflowui-polish-and-qa/SKILL.md and tighten this app for a buyer demo: smooth inputs, loading states, better spacing, mobile fixes, and more product-grade empty states.
Prepare for Databricks¶
Use skills/brickflowui-databricks-delivery/SKILL.md and make this app deployment-ready for Databricks Apps with a mock-data fallback and packaging checks.
Build a geometric concept surface¶
Use skills/brickflowui-geometric-ui/SKILL.md and adapt examples/geometric_signal_lab/app.py into a premium showcase surface that feels like a real product concept, not a dashboard disguised as one.
Best practice¶
Whenever possible, ask the assistant to adapt a real example instead of generating a new app from scratch. That keeps the output closer to the framework's proven patterns and lowers the chance of drifting into a generic web-app shape that BrickflowUI does not need.
Validation expectation¶
For serious BrickflowUI work, a good AI assistant should not stop at code generation. It should also:
- smoke-run the touched example locally
- check theme and loading behavior conceptually
- keep docs and examples aligned with the shipped API
- avoid promoting weak or fake-looking examples
- explain the tradeoffs when a design requires capabilities that are not yet first-class