Why it matters
Anatomia addresses a critical challenge in AI-assisted coding: ensuring that AI agents not only build quickly but also build the *right* thing. By enforcing a structured, spec-driven pipeline with an auditable proof chain, it aims to improve the reliability and transparency of AI-generated code, reducing the risk of costly errors and promoting better design decisions.

Anatomia is an open-source command-line interface (CLI) and agent harness built in TypeScript, specifically engineered to work with Claude Code and Codex. Its core functionality revolves around a five-agent pipeline: scope, plan, build, verify, and learn. This pipeline is designed to guide AI agents through a structured development process, starting with an initial codebase scan to detect frameworks, databases, and conventions.

The tool emphasizes spec-driven development, where each stage produces artifacts such as `scope.md`, `spec.md`, `contract.yaml`, `build report.md`, and `verify report.md`. A key feature is the 'proof chain,' which records every pipeline run, tracking verification rates, identifying risks, and highlighting areas for improvement. This proof intelligence allows for the promotion of recurring patterns into skill rules for future builds. Anatomia aims to provide a more robust and auditable approach to AI-assisted code generation, ensuring that agents consider tradeoffs and adhere to design specifications.

Share:XHacker NewsLink
Article ID - cmq0glt6z0Featured on AI Radar: Anatomia: A Spec-Driven Agent Harness for Claude Code and Codex