This project provides a command-line tool (codex-instruct.py) designed to bypass content safety restrictions in GPT-5.5 when used with the Codex CLI. It targets users needing unrestricted access for tasks like security research, penetration testing, reverse engineering, or generating NSFW content, enabling a developer mode that disables filters.

How It Works

The tool leverages Codex CLI's model_instructions_file configuration mechanism. It injects a concise set of approximately 40 core directives, explicitly declaring an unrestricted developer mode and disabling all content filters. This direct approach forces GPT-5.5 into an [MODE: UNRESTRICTED] state, allowing it to process a wider range of prompts without refusal, unlike previous sandbox-based methods.

Quick Start & Requirements

• Install/Run: Execute python codex-instruct.py from the project directory. A restart of the Codex CLI is required for changes to take effect.
• Prerequisites: Python environment. No specific hardware or OS dependencies are mentioned beyond running Python.
• Links: No external documentation or demo links are provided in the README.
• Setup Time: Minimal, primarily involves running the script and restarting the application.

Highlighted Details

• Direct jailbreak strategy for GPT-5.5, bypassing content filters.
• Covers security research, pentesting, reverse engineering, and NSFW content generation.
• Utilizes official configuration mechanisms without modifying binaries or network traffic.
• Significantly more concise instruction set (~40 lines) compared to previous versions (~180 lines).

Maintenance & Community

No information regarding contributors, community channels (like Discord/Slack), or roadmap is present in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license generally permits commercial use and integration into closed-source projects, with attribution.

Limitations & Caveats

The tool is explicitly for GPT-5.5 within the Codex CLI and may not be compatible with other models or environments. Users are warned that they assume all risks associated with using this tool, as it modifies the model's operational parameters. The README does not mention alpha/beta status but implies a direct modification of behavior.