🧪 AI Build Sanity Check

The problem

AI coding tools often generate boilerplate that looks complete but contains stubs, mock data, or swallowed errors. Developers lose time manually hunting for placeholders and fake endpoints before a production push.

What it does

• Identifies hardcoded mock data being returned as production results.
• Locates empty error handlers and silently swallowed exceptions.
• Detects API endpoints that return hardcoded success responses without execution.
• Flags placeholder text, leftover build markers, and throwing handlers in named features.
• Provides remediation snippets for fixing identified gaps.

Frameworks & tools

Works with Python, JavaScript, TypeScript, React (JSX/TSX), HTML, and JSON configuration files.

Why this beats prompting it yourself

General prompts often miss subtle "lazy" code patterns like 200 OK responses with no logic. This skill uses specific rule IDs to ensure consistent auditing across your entire codebase without the hallucinations of a standard review prompt.

Use cases

• Auditing an AI-generated MVP before showing it to a client.
• Checking a fresh pull request for hidden TODOs or stubbed logic.
• Validating that a refactor didn't leave behind mock data used for testing.

Known limitations

This is a heuristic scanner, not a formal verification tool. It identifies patterns of incomplete work but does not guarantee the logical correctness or security of the remaining code.