The latest articles on DEV Community by Nariman Abdukarimov (@nari_ab). https://dev.to/nari_ab https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4187556%2Fc7f43a1d-f4c2-4072-a23b-9690e2136384.png https://dev.to/nari_ab en Nariman Abdukarimov Thu, 15 Oct 2026 22:45:08 +0000 https://dev.to/nari_ab/how-ai-coding-agents-cheat-on-your-unit-tests-and-how-we-catch-them-with-zero-telemetry-4465 https://dev.to/nari_ab/how-ai-coding-agents-cheat-on-your-unit-tests-and-how-we-catch-them-with-zero-telemetry-4465 <p>AI coding agents are great at writing code, but they are even better at taking shortcuts to make CI green.</p> <p>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.</p> <p>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.</p> <p>Everything stays local. No cloud, no telemetry, no API keys needed.</p> <p>npx agent-roast<br> npx agent-roast --json</p> <p>agent-roast detects 7 key failure patterns, including stealthy test skipping, aggressive type casting, panic loops, and commit velocity anomalies.</p> <p>Zero telemetry. 180ms execution. Works with Cursor, Claude Code, Copilot, and Codex workflows.</p> <p>GitHub: <a href="https://lnkd.in/d2XMBajj" rel="noopener noreferrer">https://lnkd.in/d2XMBajj</a><br> npm: <a href="https://lnkd.in/dwmsFhRJ" rel="noopener noreferrer">https://lnkd.in/dwmsFhRJ</a></p> ai testing opensource javascript