Intent Layer

Intent Layer is an open source skill for AI coding agents that sets up hierarchical AGENTS.md files across a codebase. The idea is to give an agent the map a senior engineer carries in their head: what each folder owns, what breaks if it is changed incorrectly, and where the real logic lives. Each file records a folder's purpose, its contracts and its pitfalls in plain markdown. It is aimed at developers who use agents such as Claude Code, Codex, Cursor and Copilot on large repositories, where the author says agents waste tokens exploring dead ends. When run on a project, the skill detects existing CLAUDE.md and AGENTS.md files, analyzes the codebase structure, suggests where to add context nodes and asks what patterns and pitfalls should be documented. It can be run again later to audit existing nodes or find new candidates as the code grows. The author reports one example where an agent loaded 16,000 tokens instead of 40,000 and found a bug on the first attempt. The skill is published as open source by the Crafter Station group and is built on the Intent Layer method by Tyler Brandt. It is the first of several planned context engineering skills. No price is listed. The listed URL is the author's January 2026 announcement post rather than a dedicated product site.