UnvibeCode: what AI-written code actually does

Context

Teams now ship code an AI assistant wrote faster than anyone could read it. Reading files one by one doesn't show the business workflow, and pasting a whole repository into a model loses the connections between entry points, state changes, dependencies and outcomes.

The problem

Before you can fix, migrate or review an AI-built system, you need to know what it does and where each behavior lives.

What was built

• Business workflow map: execution paths, decisions, dependencies and state changes reconstructed from source
• Connected code map: interactive relationships between files, imports and symbols, with focused context you can download
• Business risk findings: implementation defects with business impact, each supported by source evidence, kept separate from cases where the evidence is insufficient
• Complete repository context: a normalized package of the repository for review or further investigation

See it

• Source on GitHub
• Package on PyPI
• Sample: smolagents
• Sample: PocketFlow
• Sample: mini-swe-agent
• Sample: django-payments