anthropics
Anti-Pattern Detection
Anti-Pattern Detection
Priority: MEDIUM
Problem: Claude Code doesn't recognize when its own behavior is unproductive. Example: creating 5 different expect scripts when this approach clearly wasn't working.
Current Behavior:
Creates multiple temporary scripts without progress Tries same command with slight variations repeatedly No self-awareness of unproductive patterns Expected Behavior:
Detect: "I've created 3 scripts without progress" Auto-suggest: "This approach isn't working. Let me step back and try a different strategy." Flag repetitive variations: "I'm trying similar approaches repeatedly. Should I research alternatives?" Suggested Patterns to Detect:
Creating >2 temporary scripts without progress Running same command 5+ times with slight variations No progress for 30+ minutes (measured by errors continuing) Circular logic (trying A → fails → try B → fails → try A again)
