AI Code Walkthrough

45% of developers say debugging AI-generated code takes longer than writing it themselves. The problem isn't that AI writes bad code — it's that developers accept code they haven't fully read, and the silent assumptions the AI made only surface later as bugs.

This skill reads AI-generated code (a block, a file, or a diff) and produces a structured walkthrough: what each section does, what assumptions the AI silently made, and what risks those assumptions introduce. Use it before you commit, before you open a PR, or whenever you look at generated code and think "I'm not sure I understand all of this."

When to use this skill

Use this skill whenever you have AI-generated code you're about to commit and want to understand it before it becomes part of your codebase. It works on any language, any framework, any size — from a single function to a full file. It does not lint or format; it explains intent, surfaces hidden assumptions, and flags risks a quick skim would miss.

Workflow

Step 1 — Collect the code

Ask what code to walk through. The user can provide it as:

• A code block pasted directly in the conversation
• A file path (read the file)
• A git diff (git diff, git diff --cached, or a specific commit range)
• A PR diff (if they provide a URL or PR number)