Literate (Extended-by-Agent) Programming
The open-source development paradigm engineered to unify human design intent with AI agent execution.
LEAP (Literate Extended-by-Agent Programming) modernizes Donald Knuth's classic literate programming concept, unifying high-fidelity documentation, strict gating structures, and progressive test-driven development into a seamless, high-integrity workflow optimized for human engineers and AI coding assistants.
The Core Principles of LEAP
Documentation-First Intent
All development begins by defining strict objectives in goals.md and detailed architectural routes in plan.md. Design decisions are committed and frozen before coding begins.
Sequential Phase Gating
Progress is verified against explicit success criteria at the end of every feature phase. This prevents compound errors and guarantees that the AI developer is always aligned with human oversight.
Rigorous TDD Cycles
Implementation is strictly driven by Test-Driven Development (TDD). Every code change is verified by an automated test suite, enforcing high-fidelity execution and 90%+ code coverage.
Unified Taxonomy
A standardized, agent-parsable Markdown namespace synchronizes intent, status, and instruction sets between the human engineer, the IDE environment, and LLM subagents.
Explore the Public Repository
The formal LEAP methodology specification, standard subdirectory templates, and the automated Markdown compliance checker are open-source and actively maintained on GitHub.
