CodeAssay

Git forensics tool for analyzing AI-authored code quality. Detects AI-generated commits, identifies rework, classifies root causes, and generates interactive dashboards.

pip install codeassay

codeassay scan /path/to/your/repo
codeassay report
codeassay dashboard

The Problem

AI coding assistants generate code fast, but how much of it actually sticks? Without tracking, you can’t distinguish between tools that nail it 90% of the time and ones that create subtle bugs you spend hours debugging. Most teams have zero visibility into their AI code quality.

How It Works

1

Detect AI Commits

Identifies AI-authored commits via Co-Authored-By trailers (Claude, Copilot, GPT), branch patterns, and manual tags.

2

Track Rework

Traces subsequent commits that modify AI-authored lines using git blame ancestry. Catches both human fixes and AI-on-AI rework.

3

Classify & Report

Categorizes rework into 7 types (bug fix, misunderstanding, style, etc.) and generates CLI reports, markdown, or interactive dashboards.

Rework Classification

Every rework event is automatically classified using commit message analysis and diff shape heuristics:

Bug Fix
Code had a defect discovered later
Misunderstanding
AI built the wrong thing
Test Failure
Didn’t pass tests on first attempt
Style/Convention
Didn’t follow project patterns
Security Issue
Introduced a vulnerability
Incomplete
Left TODOs or placeholders
Over-engineering
Unnecessary complexity that was stripped out

Features

Interactive Dashboard

Self-contained HTML file with Chart.js visualizations. Summary cards, category doughnut, trend lines, file hotspots. Works offline, shareable.

Multi-Repo Support

Scan any number of repos in one command. Repo-agnostic — no configuration needed. Just point and scan.

SQLite Storage

Data stored in .codeassay/quality.db — queryable with any SQL tool. No vendor lock-in, full data ownership.

Zero Dependencies

Python standard library only. No heavy frameworks, no external services, works completely offline.

.codeassayignore

Exclude files from analysis using gitignore-style patterns. Filter out documentation and config noise for cleaner metrics.

Claude Code Plugin

Install as a Claude Code plugin for /codeassay skill access directly in your development sessions.

Commands

Command Purpose
codeassay scan <paths> Scan repos for AI commits and rework
codeassay report Generate quality report (CLI or markdown)
codeassay dashboard Open interactive HTML dashboard
codeassay commits List AI-authored commits
codeassay rework List rework events with classification
codeassay reclassify Override a classification
codeassay export Export raw data as JSON

Get Started

Install CodeAssay and start measuring your AI code quality in under a minute.