Anticipating a common question: “Uh… you mean AGENTS.md?” Short answer: • AGENTS.md = broad, freeform guidance for any agent (already supported by Copilot).

Anticipating a common question: “Uh… you mean AGENTS.md?” Short answer: • AGENTS.md = broad, freeform guidance for any agent (already supported by Copilot). • LLMS.md = structured, schema-like spec tailored for LLMs (dirs, entry points, ignore, run/test). They’re complementary, not competing. Think of it as: • AGENTS.md → notes for agents (“how to behave here”) • LLMS.md → metadata for models (“what matters in this repo”) Why LLMS.md? We’re aligning it with llms.txt on the web. Websites have llms.txt to guide LLMs to important pages. Repos now get LLMS.md to guide LLMs to important files. Together, they form a predictable convention across ecosystems. Where tools fit in: Services like gitingest already flatten repos into an LLM-readable string. LLMS.md gives them an upstream spec so they don’t have to guess what to include or ignore. Big picture: • llms.txt → websites • LLMS.md → code repos • gitingest & friends → renderers that consume the spec • Agents/IDEs → benefit from curated, less noisy context That’s the ecosystem we’d like to help build. Feedback welcome!