Show HN: Leo, a Markdown Engineering Framework for AI Coding Agents

Summary

Leo is presented as an engineering framework for working with AI coding agents, rather than as a package or hosted service. Its main configuration is a `cursorrules` file containing 44 rules, supplemented by 133 Markdown files that define roles, canons, and protocols for enterprise-oriented development. The author says the framework is intended to support very large AI-native projects, citing a target or example of more than 300,000 lines of code and quality constraints including zero TODO items and cyclomatic complexity of no more than 10 as measured by Lizard. As an example of its use, the author says they built an AI Nodes Constructor pipeline IDE, described as similar to n8n, alone over three to four months. Reported project checks and structure include 368 schema constraints, 77 row-level-security policies, more than 120 Alembic migrations, over 3,900 tests, and more than 10,200 functions. The post links to additional documentation about the framework and to another project built with LEO, but it does not provide an independent evaluation of whether the process improves software quality or productivity.