home / skills / petekp / claude-code-setup / transparent-ui
This skill helps you generate temporary debugging interfaces that reveal internal state, transitions, and data flow for complex systems.
npx playbooks add skill petekp/claude-code-setup --skill transparent-uiReview the files below or copy the command above to add this skill to your agents.
---
name: transparent-ui
description: Create standalone debugging interfaces that reveal the internal workings of complex systems through interactive visualization. Use when the user wants to understand how something works, debug internal state, visualize data flow, see what happens when they interact with the system, or build a debug panel for any complex mechanism. Triggers on requests like "I don't understand how this works", "show me what's happening", "visualize the state machine", "build a debug view for this", "help me see the data flow", "make this transparent", or any request to understand, debug, or visualize internal system behavior. Applies to state machines, rendering systems, event flows, algorithms, animations, data pipelines, CSS calculations, database queries, or any system with non-obvious internal workings.
---
# Transparent UI
Build temporary debugging interfaces that make invisible system behavior visible. These are development-only routes/pages that reveal internal state, transitions, and data flow through interactive visualization.
## Core Principles
**Make the invisible visible.** Show state that normally exists only in memory. Reveal transitions that happen too fast to observe. Surface the "why" behind system behavior.
**Temporary by design.** These are debugging tools, not production features. Keep changes isolated for easy removal. Use dev-only routes and environment checks.
**Use existing components.** Build with the project's component library and design system. The visualization should feel native to the codebase, not like a foreign debugging tool.
**Match instrumentation to context.** Sometimes minimal logging is enough; sometimes a full observable wrapper is needed. Choose the lightest approach that captures the necessary information.
## Workflow
### Step 1: Identify What to Make Transparent
Analyze the system and identify:
1. **State**: What values change over time? What's the shape of the data?
2. **Transitions**: What events trigger state changes? What's the sequence?
3. **Relationships**: How do components/modules communicate? What depends on what?
4. **Hidden logic**: What conditions, thresholds, or rules govern behavior?
Ask the user clarifying questions if the system boundary is unclear.
### Step 2: Choose Visualization Approach
Select based on the system's nature. See [references/patterns.md](references/patterns.md) for domain-specific guidance.
| System Type | Primary Visualization | Key Elements |
|-------------|----------------------|--------------|
| State machines | Node-edge graph | States as nodes, transitions as edges, current state highlighted |
| Data flow | Directed graph or Sankey | Sources, transformations, sinks with data flowing between |
| Event systems | Timeline or sequence diagram | Events on time axis, handlers, propagation paths |
| Algorithms | Step-by-step animation | Data structure state at each step, highlighting active elements |
| Render/update cycles | Tree with diff overlay | Component tree, what re-rendered, why |
| Animations | Timeline scrubber | Keyframes, easing curves, current progress |
| CSS/Layout | Box model overlay | Computed values, constraint sources |
### Step 3: Design Interactivity Level
Layer interactivity based on debugging needs:
**Level 1 - Observation**: Real-time display of current state and recent changes. Always include this.
**Level 2 - Inspection**: Click/hover to see details. Expand nodes, view full payloads, trace data origins.
**Level 3 - Manipulation**: Trigger events, modify state, inject test data. Useful for reproducing edge cases.
**Level 4 - Time travel**: Record history, scrub through past states, replay sequences. Essential for race conditions and timing bugs.
Start with Level 1-2. Add 3-4 when the user needs them.
### Step 4: Instrument the System
Choose instrumentation strategy based on invasiveness tolerance:
**Minimal (prefer when possible)**:
- Add event emitters at key points
- Wrap state setters to broadcast changes
- Use existing debug/logging hooks if available
**Wrapper/Proxy approach**:
- Create observable wrappers around the system
- Intercept calls without modifying core code
- Useful for third-party code or when core modifications are undesirable
**Implementation patterns**:
```typescript
// Event emitter pattern - add to existing code
const debugEmitter = new EventEmitter();
function transition(from: State, to: State, event: string) {
debugEmitter.emit('transition', { from, to, event, timestamp: Date.now() });
// ... existing logic
}
// Proxy pattern - wrap without modifying
function createObservableStore<T>(store: T): T & { subscribe: (fn: Listener) => void } {
const listeners: Listener[] = [];
return new Proxy(store, {
set(target, prop, value) {
const oldValue = target[prop];
target[prop] = value;
listeners.forEach(fn => fn({ prop, oldValue, newValue: value }));
return true;
}
});
}
```
### Step 5: Build the Debug Route
Create a development-only route that:
1. **Guards against production**: Check `process.env.NODE_ENV === 'development'`
2. **Connects to instrumentation**: Subscribe to events/state changes
3. **Renders visualization**: Use the project's components where possible
4. **Provides controls**: Play/pause, speed, filters, time scrubbing as needed
**Route structure** (Next.js example):
```
app/
__dev/
transparent/
[system]/
page.tsx # Dynamic route for different systems
```
Or simpler:
```
app/
__dev/
state-machine/
page.tsx
```
**Recommended libraries** (install only if not already in project):
- **react-flow** or **reactflow**: Node-edge graphs for state machines, data flow
- **framer-motion**: Smooth transitions in visualization itself
- **Existing charting library**: If project already has one, use it
### Step 6: Document Removal Path
At the top of every file created, add:
```typescript
/**
* TRANSPARENT-UI DEBUG TOOL
*
* Temporary debugging visualization. Remove when no longer needed:
* 1. Delete this file: app/__dev/[name]/page.tsx
* 2. Delete instrumentation: src/lib/[system]-debug.ts
* 3. Remove debug hooks from: src/lib/[system].ts (lines XX-YY)
*
* Created for: [description of what this helps debug]
*/
```
## Cleanup
When the user asks to remove the transparent UI or is done debugging:
1. **Delete debug route**: Remove the `__dev/` page(s)
2. **Remove instrumentation**: Delete event emitters, proxies, debug hooks
3. **Uninstall unused deps**: If visualization libraries were added solely for this
4. **Verify no remnants**: Search for `debugEmitter`, `TRANSPARENT-UI`, or similar markers
Provide a summary of removed files and modified lines.
## Domain-Specific Patterns
For detailed visualization patterns, layouts, and code examples organized by system type, see [references/patterns.md](references/patterns.md).
This skill creates temporary, development-only debugging interfaces that make invisible system behavior visible through interactive visualizations. It helps developers inspect state, trace transitions, and explore data flow for complex systems like state machines, render cycles, event pipelines, and algorithms. The interfaces are designed to be easy to add and remove and to integrate with the project’s existing components and tooling.
I identify the parts of the system to expose (state, transitions, relationships, hidden logic), pick an appropriate visualization (graphs, timelines, tree diffs, or step animations), and instrument the system with the lightest-touch approach that captures needed information. The skill wires those instrumentation points to a dev-only route or page that subscribes to events, renders real-time views, and provides inspection and manipulation controls. It also documents the files and hooks to remove when debugging is finished.
Will this change production behavior?
No. The debug routes and instrumentation are intended to be dev-only and guarded by environment checks; use wrappers and emitters so core logic remains unchanged.
How invasive is the instrumentation?
Prefer minimal options: add event emitters or wrap stores with proxies. Only modify core code when strictly necessary and document changes for removal.