prompt-injection-detect

Detect prompt injection patterns in user inputs and external data sources.

1|1|Updated Apr 25, 2026
One-click install
npx skills add https://github.com/chenchunrun/onyx-soc --skill prompt-injection-detect
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: prompt-injection-detect
Source: https://github.com/chenchunrun/onyx-soc/tree/main/skills/prompt-injection-detect
Command: npx skills add https://github.com/chenchunrun/onyx-soc --skill prompt-injection-detect

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Detect prompt injection threats and defend AI systems.

Core Features & Use Cases

  • Real-time detection of direct prompt injection patterns in user inputs.
  • Indirect prompt injection detection from external sources including webpages and documents.
  • Multilingual coverage and encoding-based pattern recognition to identify obfuscated prompts.
  • Integrated risk scoring and actionable defense recommendations for operator workflows.

Quick Start

Provide a text sample for analysis and receive detected injection patterns with a risk assessment.

Frequently Asked Questions about prompt-injection-detect

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

FAQPage Schema
How do I detect prompt injection attacks in AI systems?

To detect prompt injection attacks, you provide a text sample for analysis to identify malicious patterns. The system applies pattern matching, multilingual checks, and encoding decoding to return detected injection patterns with a risk assessment.

Can I check external web pages and documents for indirect prompt injection?

Yes, indirect prompt injection detection from external sources including web pages and documents is supported. It analyzes these data inputs to identify hidden malicious instructions and provides actionable defense recommendations.

Does prompt injection detection work with multilingual and encoded text?

Prompt injection detection works with multilingual coverage and encoding-based pattern recognition. This identifies obfuscated prompts by decoding encoded text, ensuring threats across different languages and formats are caught.

How do I assess the risk level of detected prompt injection threats?

You assess risk levels using the integrated risk scoring feature. After analyzing text samples, the system generates a quantitative risk score alongside actionable defense recommendations to secure operator workflows.

What is the best way to defend against direct prompt injection in multi-turn conversations?

The best way to defend against direct prompt injection in multi-turn conversations is applying real-time detection to each turn. This analyzes ongoing user inputs to identify and score injection patterns immediately.

When should I use automated injection detection instead of manual review?

You should use automated injection detection when processing high volumes of single-turn prompts, multi-turn conversations, or external data inputs. Manual review cannot efficiently decode encoded patterns or perform multilingual checks at scale.