skills.rest
Supporter09
Supporter09COMMUNITY
Explore all skills by this author →
GitHub

prompt-guard

Detect prompt injections and jailbreak attempts in LLM inputs.

Updated May 4, 2026
Search Tags:#multilingual#prompt injection#content filtering#jailbreak detection#llm security#rag filtering
One-click install
npx skills add https://github.com/Supporter09/Face_Anti_Spoofing_Biometric --skill prompt-guard-supporter09
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prompt-guard
Source: https://github.com/Supporter09/Face_Anti_Spoofing_Biometric/tree/main/.claude/skills/prompt-guard
Command: npx skills add https://github.com/Supporter09/Face_Anti_Spoofing_Biometric --skill prompt-guard-supporter09

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prompt Guard helps teams detect prompt injections and jailbreak attempts before they can manipulate an LLM, reducing the risk of unsafe instructions, policy bypasses, and poisoned context.

Core Features & Use Cases

  • User prompt filtering: Score incoming messages and block suspicious jailbreak attempts before generation.
  • Third-party data screening: Inspect API responses, web content, and retrieved documents for embedded instructions or context poisoning.
  • Batch RAG protection: Process large document sets efficiently to separate safe retrievals from malicious ones.
  • Use case: A chatbot can screen both user prompts and retrieved documents, then only pass low-risk text into the model.

Quick Start

Use Prompt Guard to score the provided text for jailbreak risk and return whether it should be allowed into the LLM context.

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 injections and jailbreak attempts in user prompts before sending them to an LLM?▼

You detect prompt injections by scoring user prompts with a lightweight transformer classifier that assigns BENIGN, INJECTION, or JAILBREAK labels to block malicious inputs before LLM generation.

What is the best way to filter retrieved documents for context poisoning in a RAG workflow?▼

Batch RAG filtering screens retrieved documents by applying threshold-based scoring to separate safe retrievals from malicious embedded instructions, preventing context poisoning.

Does prompt injection detection work with multilingual inputs?▼

Yes, multilingual prompt injection detection applies to user prompts, retrieved documents, and API responses across enterprise and research workflows to identify jailbreak attempts.

Can I use threshold-based scoring to block jailbreak attempts in third-party API responses?▼

Yes, third-party data screening applies threshold-based scoring to API responses and web content to detect embedded instructions, blocking suspicious jailbreak attempts before they reach your LLM.

How do I implement low-latency jailbreak detection for high-volume chatbot conversations?▼

Low-latency jailbreak detection uses a lightweight transformer classifier designed for rapid threshold-based scoring of user prompts and retrieved documents in high-volume chatbot workflows.

Related Skills

Explore More Software Engineering Skills→
ricablericable

prompt-guard

Detect prompt injections and jailbreak attempts in LLM inputs.

Community
Intermediate
cardmagiccardmagic

classifier

Classify text sentiment and spam using pre-trained CLI models.

Community
Intermediate
ThedouglerThedougler

ai-engineer

Develop LLM applications, RAG systems, and agents with vector search and agent frameworks.

Community
Advanced
RazisafirRazisafir
1

detecting-ai-model-prompt-injection-attacks

Detect prompt injection attacks in LLM inputs using regex, heuristics, and DeBERTa classification.

Community
Advanced