Vibe-Code Smells Catalog
15 anti-patterns that indicate technical debt in AI-assisted development.
Ambiguous Goal
Ambiguous Goal: A vague prompt that forces AI to make architectural decisions. Learn how to prevent this fundamental AI coding mistake.
Prompt Drift
Prompt Drift: Successive chat iterations cause AI to lose track of original intent. Learn to detect and fix this common AI coding anti-pattern.
Context Flood
Context Flood: Dumping too much unstructured code into AI causes degraded reasoning. Learn the 'Curated Context' strategy.
Over-Delegation
Over-Delegation: Giving AI agents too much autonomy without supervision. The fastest way to technical bankruptcy.
No Boundaries
No Boundaries: Allowing AI agents unrestricted access to your codebase. Learn how to set spatial and operational boundaries.
Magic Black Box
Magic Black Box: Working code that no one understands. The ultimate driver of comprehension debt in AI projects.
Hidden Dependencies
Hidden Dependencies: Implicit coupling created by AI between distant modules. Learn to detect and fix invisible threads in your AI code.
Over-Structuring
Over-Structuring: AI creates unnecessary abstractions and complexity. Learn to apply YAGNI principles to AI-generated architecture.
Style Drift
Style Drift: Inconsistent code style and patterns across AI-generated modules. Learn to maintain a unified team voice with AI.
Missing Self-Review
Missing Self-Review: Merging AI code without understanding it. Learn why 'Tests Pass, Ship It' is a dangerous AI mantra.
Copy-Paste Loops
Copy-Paste Loops: Repeatedly adding AI snippets without checking for existing logic. Learn to prevent duplication debt.
Over-Iteration
Over-Iteration: Endless chat 'fixes' that make code worse. Learn the '3-5 Turn Rule' to maintain AI quality.
Circular Testing
Circular Testing: AI writing both code and tests results in false confidence. Learn to write effective AI validation strategies.
Silent Failure
Silent Failure: AI-generated code that swallows errors and hides production issues. Learn to build observable AI systems.
Ethical Blindspot
Ethical Blindspot: AI code that violates privacy, accessibility, or legal norms. Learn to maintain responsibility in AI development.
How to Use the Catalog
Count how many of these smells apply to your current codebase. 0-2 is healthy, 3-5 indicates concerning debt, and 6+ means your project is in crisis.
Get Remediation Templates