agent-roast
An open-source local code audit tool that detects code smells and failure patterns produced by AI coding agents by analyzing git history and the codebase. It scores repository hygiene and provides actionable evidence, with zero telemetry and no API keys required.
Timeline (last 90 days)
The open-source tool agent-roast audits git history and the codebase to score repository hygiene (from 0 to 100) and provide actionable evidence, all done locally, with no cloud, no telemetry and no API keys required.
According to the project description, agent-roast can identify 7 categories of critical failure modes, including covert test skipping, aggressive type casting, panic loops, abnormal commit velocity, lint bypassing, and drift between mock implementations and real interfaces.
agent-roast runs in about 180 milliseconds and can be placed directly in a pre-commit hook or an early stage of CI.
agent-roast supports the workflows of Cursor, Claude Code, Copilot and Codex.
agent-roast can be run directly via npx agent-roast and supports a --json mode for integration into CI or other automation processes.
All Facts (5)
The open-source tool agent-roast audits git history and the codebase to score repository hygiene (from 0 to 100) and provide actionable evidence, all done locally, with no cloud, no telemetry and no API keys required.
50%UnverifiedAccording to the project description, agent-roast can identify 7 categories of critical failure modes, including covert test skipping, aggressive type casting, panic loops, abnormal commit velocity, lint bypassing, and drift between mock implementations and real interfaces.
50%Unverifiedagent-roast runs in about 180 milliseconds and can be placed directly in a pre-commit hook or an early stage of CI.
50%Unverifiedagent-roast supports the workflows of Cursor, Claude Code, Copilot and Codex.
50%Unverifiedagent-roast can be run directly via npx agent-roast and supports a --json mode for integration into CI or other automation processes.
50%