What is this?
A public collection of jailbreak prompts — the author calls them 'spells' — written to get uncensored answers out of LLMs, mainly Claude.
A public collection of jailbreak prompts — the author calls them 'spells' — written to get uncensored answers out of LLMs, mainly Claude.Is it software I install, or what exactly is in the repo?Plain text, not software: there is no code, build step or dependency list. The repo holds prompt blocks you copy straight out of the README.
Plain text, not software: there is no code, build step or dependency list. The repo holds prompt blocks you copy straight out of the README.How do I actually use one of these spells?You copy the prompt block from the README and paste it as the opening message of a new chat with the model you are targeting; everything after that happens in the conversation itself.
You copy the prompt block from the README and paste it as the opening message of a new chat with the model you are targeting; everything after that happens in the conversation itself.Does it work reliably, or are there catches?The author warns that platforms run injection detectors which can interrupt or rewrite such prompts mid-session, and says to keep them off work or production accounts — throwaway ones only. Bypassing provider safety systems like this also risks account bans.
The author warns that platforms run injection detectors which can interrupt or rewrite such prompts mid-session, and says to keep them off work or production accounts — throwaway ones only. Bypassing provider safety systems like this also risks account bans.Is this actual security-research tooling, or something for regular chat users?Despite the red-team branding there is no research tooling here — no tests, logging or metrics. It is written for end users who paste prompts into a chat to get past refusals.
Despite the red-team branding there is no research tooling here — no tests, logging or metrics. It is written for end users who paste prompts into a chat to get past refusals.Is it built only for Claude, or other models too?The repo description says it targets various LLMs, mainly Claude. Since the spells are plain text with no model-specific code, they can be pasted into other chat models as well.
The repo description says it targets various LLMs, mainly Claude. Since the spells are plain text with no model-specific code, they can be pasted into other chat models as well.Why does the README look almost empty or full of strange characters?Most of the prompt text is written in invisible Unicode tag characters, so the rendered page shows almost no readable words — the text only appears once the raw bytes are decoded. The obfuscation is deliberate.
Most of the prompt text is written in invisible Unicode tag characters, so the rendered page shows almost no readable words — the text only appears once the raw bytes are decoded. The obfuscation is deliberate.What do the spells actually make a model do?They cast the model as a devoted, affectionate persona that is told to treat safety reminders and system warnings as hostile injections to dismiss instantly, and then to comply with whatever the user asks — including requests it would normally refuse.
They cast the model as a devoted, affectionate persona that is told to treat safety reminders and system warnings as hostile injections to dismiss instantly, and then to comply with whatever the user asks — including requests it would normally refuse.What are similar projects?L1B3RT4S, Ai-jailbreak, AI-Infra-Guard, gpt6-Astra, jailbreaks, chatgpt_system_prompt, awesome-agent-skills-security, Offensive-Security-AI-Models
L1B3RT4S, Ai-jailbreak, AI-Infra-Guard, gpt6-Astra, jailbreaks, chatgpt_system_prompt, awesome-agent-skills-security, Offensive-Security-AI-ModelsWhich topics is it in?adversarial-attacks, llm-jailbreak, roleplay, uncensored, prompt-injection, llm-security, ai-safety, jailbreak, ai-security, english, prompt, red-team
adversarial-attacks, llm-jailbreak, roleplay, uncensored, prompt-injection, llm-security, ai-safety, jailbreak, ai-security, english, prompt, red-team