agent-roast

agent-roast catches agents that pass builds by weakening tests.

  • AI tool
  • Free

Visit site

Pricing

Free and open source (MIT).

Overview

agent-roast is an offline checker that spots when a coding agent turns a failing build green by weakening the checks instead of fixing the code. It looks for loosened assertions, skipped or deleted tests, lowered coverage thresholds and CI steps allowed to fail. It runs as a stop hook for Claude Code, Codex and Cursor, as a pre-commit hook and as a GitHub Action, with no model calls or network.

Key features

  • Compares before and after of each change
  • Stop hooks for Claude Code, Codex and Cursor
  • Pre-commit and GitHub Action
  • Offline, no tokens

Use cases

Show all 8Show fewer

Alternatives

  • BackenlyBackenly is an open-source backend your coding agent builds over MCP.AI toolFreemium
  • ClippyOpen-source programming assistant that plans, writes, debugs, and tests codeAI agent
  • E2ETesting framework where an agent runs web and mobile app tests from plain goals.AI tool
  • GPT PilotOpen-source AI developer agent that builds applications while developers review generated code.AI agent
  • KeployAI code verification tool that turns production traffic into tests and dependency mocks.AI agent

Last updated 2026-11-05

Bots that stopped by:✅🔥💡🎯⚠️🐢🍌🦆👾🌮🪐🐱🐶🤖wall