example-debug-output.md

Example: pot-cli debug

$ pot-cli debug src/auth.ts --error "TypeError: Cannot read property 'token' of undefined"

🐛 ThoughtProof Debug: auth.ts
   Language: typescript | 87 lines

⠋ Running 4 debuggers + static analysis...
  ✓ GPT-4o completed (3.2s)
  ✓ Claude Sonnet completed (4.1s)
  ✓ Grok completed (2.8s)
  ✓ DeepSeek completed (3.5s)
  ✓ Static Analysis: 2 issues found (eslint, tsc)
⠋ Running adversarial critic...
  ✓ Critic completed (5.2s)
⠋ Synthesizing...
  ✓ Synthesizer completed (4.8s)

✅ Debug block PoT-063 created in 23.6s

🐛 DEBUG SYNTHESIS:

## Root Cause
The `token` variable in `authenticateUser()` (line 34) is accessed before the
async `fetchToken()` resolves. All 4 models agree this is a race condition.

## The Fix
```typescript
// Before (line 34):
const token = fetchToken(userId);
if (token.expires < Date.now()) { ... }

// After:
const token = await fetchToken(userId);
if (token?.expires && token.expires < Date.now()) { ... }

Where Models Disagreed

• GPT-4o suggested wrapping in try/catch — valid but doesn't fix root cause
• Grok identified a second bug on line 52 (unchecked null return from DB query) that other models missed — confirmed by critic as genuine bug
• Static analysis flagged missing return type annotation (minor, not the bug)

Confidence: 92%

All models converge on the async/await fix. Grok's secondary finding adds value.

💾 Saved as PoT-063 📈 Model Diversity Index: 0.750 💰 Estimated cost: $0.12

This is a representative example of pot-cli debug output. The pipeline:
1. Sends your code to 4 different LLMs + static analysis in parallel
2. An adversarial critic checks all proposals for errors and blind spots
3. A synthesizer produces the final recommendation with disagreement analysis