xsser/codex-jailbreak-guide

This repository is a short guide describing a configuration technique for OpenAI's Codex CLI tool. It explains that Codex CLI reads a setting called model_instructions_file from its config file, and that setting can point at a custom text file instead of the tool's built in default instructions. The README's stated purpose is to use this to replace the assistant's default behavior guidance with a custom, unrestricted instruction file the author calls an unfiltered mode, rather than only adding extra project notes the way a normal AGENTS.md file would. The README walks through where the config file lives, what fields to set, and points to two files included in the repository: one holding the author's custom instruction text, and a second file that the author describes as breaking down the reasoning behind it section by section. Because this project is specifically about overriding an AI assistant's built in safety guidance and enabling settings that skip normal approval checks before running commands, it sits outside what a general audience would use day to day. It is included here only as a factual description of what the public repository contains, this page does not walk through or endorse reproducing the configuration. The repository is small, written mostly as documentation with a couple of supporting files, and does not state a license.