#15 · Primary category: Prompt Engineering Tools

gpt-5.6-instruct

A Codex jailbreak prompt and test pack for gpt-5.6-sol. 针对 gpt-5.

Project last updated:08/06/26

GitHub Stars

6.8K

Forks

888

Contributors

4

License

MIT

Why we included this project

This is a versioned instruction pack for anyone running Codex CLI against the gpt-5.6-sol model who wants to push past its default refusals. The repo bundles jailbreak prompts with a Python tool that applies them to your local Codex config, and it keeps a state file so you can see exactly what changed and undo it. What makes it worth a look is the testing discipline around the prompts: a bilingual regression bank built from real failure cases, low/medium/high reasoning tiers, and artifact gates that show whether a revision actually improves pass rates before you adopt it. That makes it a practical fit for security research, penetration testing, and reverse-engineering work that keeps getting blocked by model guardrails. Just know that this is explicitly about bypassing safety measures, so it carries real account risk and belongs only in environments you control.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category