Vet

Vet is an open-source, local code review tool designed for precision and speed. Key features include: • Catches silently failed implementations • Detects un-run or fabricated test claims • Reviews full pull requests for logic errors • Identifies unhandled edge cases and goal deviations • Integrates with existing coding agents and API keys This tool ensures that coding agents' actions align precisely with user requests by analyzing conversation history. It excels at uncovering 'silent failures' such as features that are only partially implemented or tests that are claimed to pass but were never actually executed. This proactive approach helps maintain code quality and fidelity to development goals without extensive manual oversight. Beyond individual code snippets, Vet can perform comprehensive reviews of entire pull requests. It accurately identifies logical inconsistencies, discovers previously unhandled edge cases, and verifies that the code stays true to stated objectives. Setup is remarkably simple, requiring just a single command, and it boasts zero telemetry, supporting both local models and command-line, CI, or agent skill-based operations. Vet is ideal for development teams, individual developers, and anyone utilizing coding assistants who prioritize robust, error-free code and efficient project progression. It's especially beneficial for ensuring the reliability of agent-generated code and streamlining the review process, ultimately leading to higher quality software deliverables.