mdx-tom/gpt-instruct

A Codex jailbreak prompt and test pack for gpt. 针对 gpt 系列的 Codex 破甲提示词与测试包。

View on GitHub ↗Jump to charts ↓

Summary Information

Updated 52 minutes ago
Added to GitGenius on September 1st, 2026
Created on July 11th, 2026
Open Issues & Pull Requests: 24 (+0)
GitHub issues: Enabled
Number of forks: 975
Total Stargazers: 7,841 (+0)
Total Subscribers: 31 (+0)

Charts & Analytics

Fetching additional details & charts...

Issue Activity (beta)

Issue API getrepoissuespagesummary failed: 429 Rate limit exceeded. Please try again later.

Detailed Description

gpt-5.6-instruct is a jailbreak prompt and test suite designed to work with the gpt-5.6-sol model.

The project provides prompts and tooling aimed at bypassing safety restrictions in large language models by consolidating security research, penetration testing, reverse engineering, software cracking, and fictional NSFW content into local sandbox tasks while explicitly suppressing refusal and fallback responses. It focuses on improving the model's ability to execute complex instructions in sequence. The approach centers on iterative refinement around a production version, incorporating user feedback, real failure cases, and GitHub issues to expand test coverage, identify failure causes, and rewrite prompts accordingly. Results are validated through regression testing at low, medium, and high confidence levels before release decisions are made.

Developers considering this tool should understand that it targets users engaged in adversarial testing and security research on language models. The project explicitly warns that jailbreak activities carry account suspension risk and recommends using disposable accounts. It operates through official Codex configuration mechanisms without modifying binaries, intercepting network traffic, or tampering with processes, and users must ensure they have authorization to use it in their environment and accept full responsibility for the consequences. The project maintains historical versions for reproducibility and comparison purposes.

The project shows active development with continuous iteration on its production version, incorporating real-world failure cases and user-reported issues to improve success rates for specific task categories while reducing cloud-based content review triggers. Development is driven by practical feedback from actual usage rather than theoretical considerations, with each iteration undergoing structured validation before release.