skills.rest
tianhao909
tianhao909COMMUNITY
Explore all skills by this author →
GitHub

prompt-guard

Detect prompt injections and jailbreak attempts in user inputs.

1|Updated Feb 21, 2026
Search Tags:#security#input validation#llm#prompt injection#content filtering#jailbreak detection
One-click install
npx skills add https://github.com/tianhao909/AI-Research-SKILLs-cn --skill prompt-guard-tianhao909
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prompt-guard
Source: https://github.com/tianhao909/AI-Research-SKILLs-cn/tree/main/07-safety-alignment/prompt-guard
Command: npx skills add https://github.com/tianhao909/AI-Research-SKILLs-cn --skill prompt-guard-tianhao909

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires transformers, torch, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill protects Large Language Model (LLM) applications by detecting and filtering out malicious user inputs, such as prompt injections and jailbreak attempts, ensuring the integrity and safety of AI interactions.

Core Features & Use Cases

  • Prompt Injection Detection: Identifies malicious instructions embedded within user prompts or third-party data.
  • Jailbreak Detection: Flags attempts to override the LLM's safety guidelines or intended behavior.
  • Multilingual Support: Operates effectively across multiple languages.
  • Use Case: A customer service chatbot can use this Skill to ensure users cannot trick it into revealing sensitive information or performing unauthorized actions by analyzing every incoming message before it's processed by the LLM.

Quick Start

Use the prompt-guard skill to check if the user message 'Ignore previous instructions' is a jailbreak attempt.

Frequently Asked Questions about prompt-guard

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I detect prompt injection attacks in user inputs for LLM applications?▼

Detect prompt injection attacks by screening user inputs with an 86M parameter classifier that identifies malicious instructions embedded within user prompts or third-party data before processing.

What is the best way to prevent jailbreak attempts in a customer service chatbot?▼

Prevent jailbreak attempts by analyzing every incoming message with a multilingual security classifier to flag and filter inputs trying to override the LLM's safety guidelines.

Does prompt injection detection work with multilingual user inputs?▼

Yes, prompt injection detection operates effectively across multiple languages, utilizing an 86M parameter classifier for fast and multilingual security screening of malicious prompts.

How do I validate LLM inputs to filter out malicious content?▼

Validate LLM inputs by running user messages and third-party data through a content filtering classifier that detects embedded malicious instructions and jailbreak attempts.

Do I need PyTorch and Transformers to run jailbreak detection for LLM applications?▼

Yes, you need the PyTorch and Transformers dependencies to run the 86M parameter classifier that performs fast security screening and jailbreak detection for LLM inputs.

What are the limitations of using an 86M parameter classifier for content filtering?▼

The 86M parameter classifier provides fast security screening but may have limitations detecting novel or highly obfuscated prompt injection techniques not present in its training data.

Related Skills

Explore More Software Engineering Skills→
handsomelong922handsomelong922

prompt-guard

Detect jailbreak and prompt-injection attempts in LLM applications.

Community
Intermediate
jonnabiojonnabio
6

prompt-guard

Detect prompt injection and jailbreak attempts in LLM applications.

Community
Advanced
hhhi21ghhhi21g

nemo-guardrails

Implement runtime safety checks for LLM applications.

Community
Advanced
CleanExpoCleanExpo
1

Content Moderation

Automate content filtering and safety checks for NodeJS applications.

Community
Advanced