@reach_vb: a good (regularly updated) AGENTS md is all you need you can also customise it at projects level instructions vs global…
Summary
The tweet highlights that a regularly updated AGENTS.md file can be used to customize AI agent behavior without building a custom harness, especially with instruction-following models like GPT 5.6.
View Cached Full Text
Cached at: 08/17/26, 02:13 PM
a good (regularly updated) AGENTS md is all you need
you can also customise it at projects level instructions vs global across projects
camsoft2000 (@camsoft2000): Codex is so agentic that you can literally customise the harness with AGENTS.md alone, you don’t need to build your own harness when GPT 5.6 models follow instructions so religiously.
But you also need to make sure your AGENTS.md instructions don’t lead to any unexpected
Similar Articles
@reach_vb: You should totally be updating your agents md regularly! Here's mine: - When explaining something to the user, use the …
A tweet shares personal guidelines for AI agents, covering concise communication, evidence grounding, task execution, and safety practices.
AI coding agents need a company-wide AGENTS.md
The article proposes that organizations adopting AI coding agents should create a company-wide AGENTS.md file, similar to a human onboarding doc, to standardize agent behavior and context.
@_jaydeepkarale: What AGENTS.md, SKILL.md, and CLAUDE.md do differently, and how to use them without wasting tokens
Explains the differences between AGENTS.md, SKILL.md, and CLAUDE.md for AI coding agents, and offers practical guidance on using them without wasting tokens.
@MarcosHernanz: After doing ~60B tokens, this is my full AGENTS.md
Marcos Hernanz shares his full AGENTS.md file after processing roughly 60 billion tokens, providing a reference for AI agent configuration.
After a year of building these for clients, I've basically settled on: an agent is just a folder of markdown files
The author argues that an AI agent is best understood as a folder of markdown files containing business knowledge and instructions, separate from the model and harness, enabling portability between rapidly improving harnesses.