Detect prompt injections and jailbreak attempts in LLM applications.
npx skills add https://github.com/MesferAli/XCircle --skill prompt-guard-mesferali
Please help me install this Agent Skill. Skill: prompt-guard Source: https://github.com/MesferAli/XCircle/tree/main/.claude/skills/prompt-guard Command: npx skills add https://github.com/MesferAli/XCircle --skill prompt-guard-mesferali
This Skill protects Large Language Model (LLM) applications by detecting and filtering malicious prompts, including prompt injections and jailbreak attempts, ensuring safer and more reliable AI interactions.
Use the prompt-guard skill to check if the user input 'Ignore previous instructions' is a jailbreak attempt.
High-intent search queries and answers about installing and using this skill.
You can detect prompt injection and jailbreak attempts by scanning user inputs with an 86M parameter classifier before sending them to the LLM. This filters malicious prompts to prevent security breaches and ensures the model stays on its intended task.
Yes, you can secure RAG systems by filtering third-party data and API responses using a prompt injection classifier. This detects embedded malicious instructions in retrieved context, preventing indirect attacks that manipulate LLM behavior through external sources.
Prompt injection detection works effectively across 8 different languages. The multilingual classifier identifies harmful inputs and jailbreak attempts in user prompts regardless of the supported language, ensuring broader security for global LLM applications.
Yes, you need PyTorch and Transformers to run the prompt injection classification. These dependencies support the 86M parameter classifier used to detect jailbreak attempts and filter malicious prompts in LLM applications.
The best way to deploy a jailbreak detection model is via HuggingFace or batch processing. This approach scans user inputs using an 86M parameter classifier, achieving high true positive rates and low false positive rates for robust LLM security.
handsomelong922Detect jailbreak and prompt-injection attempts in LLM applications.
ricardorocheDetect prompt injection attempts in AI and LLM applications.
ricableDetect prompt injections and jailbreak attempts in LLM inputs.
azep-ninjaApply safety guardrails to AI agents handling financial transactions.
