White Room: Deepen

Push current feature through the next depth level

Take the current feature and push it deeper through the depth levels.

Depth Levels

Level 1: WHAT (Skeleton)
"Users can place bets"

Level 2: HOW (Anatomy)
"User selects line → enters stake → confirms →
 system validates balance → submits to API →
 shows confirmation or error"

Level 3: WHAT-IF (Physiology)
"If API times out after 5s: show 'checking status' →
 retry 2x with 2s delay → if still fails: show
 'bet may have been placed, checking...' →
 poll /bet-status every 3s for 30s →
 if confirmed: show success →
 if not found: show 'bet not placed, try again'"

Step 1: Identify Current Feature

Read .white-room/SESSION-STATE.json and get current_focus.feature.

If no feature is focused, ask:

Which feature would you like to deepen?
[List features with their current depth level]

Step 2: Determine Current Depth

Check the feature's status:

• •If skeleton.complete = false → at Level 0
• •If anatomy.complete = false → at Level 1
• •If physiology.complete = false → at Level 2
• •If all complete → already at max depth

Step 3: Apply Depth Drill

Level 0 → Level 1 (Skeleton)

Ask:

• •What does this feature do?
• •Who uses it?
• •Why is it needed?
• •What's in scope for this feature?

Level 1 → Level 2 (Anatomy)

Ask:

• •What are the components?
• •What data does it need?
• •What APIs does it call?
• •What does the UI look like?
• •How do the pieces connect?

Level 2 → Level 3 (Physiology)

Apply the full Depth Drill (from ~/.claude/plugins/local-marketplace/white-room/engine/depth-drill.md):

For EACH operation in the feature:

1. •

   Network Failures

   • •What if the network fails mid-operation?
   • •What's the timeout?
   • •Retry strategy?
   • •Final error message?
2. •

   Double Actions

   • •What if user clicks twice?
   • •Is it idempotent?
   • •How prevent duplicates?
3. •

   Unexpected Data

   • •What if API returns garbage?
   • •What if required fields missing?
   • •Validation strategy?
4. •

   Timing Issues

   • •What if it takes 10x longer?
   • •Loading state?
   • •Can user cancel?
5. •

   Data Integrity

   • •What if data is malformed?
   • •Size limits?
   • •Security (XSS, injection)?
6. •

   Permission Changes

   • •What if auth expires?
   • •How handle 401/403?

Step 4: Document the Depth

Update the appropriate artifact:

• •Level 1 → Update 03-FEATURES.md
• •Level 2 → Update 05-ARCHITECTURE.md, 06-DATA-MODEL.md, 07-API-CONTRACTS.md
• •Level 3 → Update 04-USER-FLOWS.md, 09-SECURITY.md

Step 5: Update State

Update SESSION-STATE.json:

{
  "features": {
    "[feature_name]": {
      "physiology": {
        "complete": false,
        "progress": 0.75  // Increased
      }
    }
  }
}

Step 6: Report Progress

═══════════════════════════════════════════════════════════════
                FEATURE DEEPENED
═══════════════════════════════════════════════════════════════
Feature: [name]
Previous Depth: Level 2 (Anatomy)
Current Depth: Level 3 (Physiology) - 75%

Completed:
✓ Network failure handling
✓ Double-click prevention
✓ Unexpected data handling

Remaining:
○ Timing/loading states
○ Data validation rules
○ Permission handling

Continue deepening? [Y/n]
═══════════════════════════════════════════════════════════════

Auto-Continue

If user doesn't say stop, continue with remaining depth drill items until the feature reaches 100% at current level.

Key Rules

1. •One question at a time - Don't overwhelm
2. •Be specific - "What timeout?" not "How handle errors?"
3. •Document as you go - Update artifacts immediately
4. •Track progress - Update SESSION-STATE.json after each item