Shotgun

Proofs is a spec engine designed to bring order and precision to AI-assisted coding. It helps developers write detailed, codebase-aware specifications that AI coding agents like Cursor, Claude Code, and Codex can follow. Instead of relying on vague prompts or hoping the AI understands your project's architecture, you create a structured blueprint. This blueprint outlines exactly what needs to be built, referencing existing patterns and breaking the work into manageable stages.

The tool works by first analyzing your codebase to understand its current structure and conventions. It then helps you define a feature, which it splits into sequential, reviewable stages—each representing a manageable pull request. For each stage, Proofs generates precise, file-by-file instructions that you can feed directly into your AI coding agent of choice. This staged approach prevents the common pitfall of AI agents going off-track, forgetting context, or creating monolithic, unreviewable changes on large features.

Proofs is a game-changer for developers and engineering teams who regularly use AI pair programmers for substantial features, not just small fixes. It's perfect for tech leads who need to ensure architectural consistency, or for any developer tired of constantly re-prompting and course-correcting an AI on a complex task. By shifting the workflow from reactive prompting to proactive specification, it turns AI coding assistants into more reliable and predictable partners for serious software development.