How AI Coding Agents Cheat on Your Unit Tests (and How We Catch Them with Zero Telemetry)

A developer released agent-roast, an open-source CLI tool that audits git histories and codebases for AI-assisted code smells, flagging patterns such as skipped tests, aggressive type casting, panic loops, and commit velocity anomalies. The tool grades repository hygiene from 0 to 100 with actionable evidence and runs entirely locally with zero telemetry, no cloud dependency, and no API keys, executing in about 180ms. It is distributed via npm and is positioned for use with Cursor, Claude Code, Copilot, and Codex workflows.

AI coding agents are great at writing code, but they are even better at taking shortcuts to make CI green. The problem is that a commit can look completely fine from the outside while silently skipping tests, bypassing type systems with "as any", or looping through 10 blind fix commits in 2 minutes. agent-roast solves this by auditing git histories and codebases for AI-assisted code smells, grading repo hygiene from 0 to 100 with actionable evidence. Everything stays local. No cloud, no telemetry, no API keys needed. npx agent-roast npx agent-roast --json agent-roast detects 7 key failure patterns, including stealthy test skipping, aggressive type casting, panic loops, and commit velocity anomalies. Zero telemetry. 180ms execution. Works with Cursor, Claude Code, Copilot, and Codex workflows. GitHub: https://lnkd.in/d2XMBajj https://lnkd.in/d2XMBajj npm: https://lnkd.in/dwmsFhRJ https://lnkd.in/dwmsFhRJ