prompt-injection-guard

Detect and block prompt-injection attempts by enforcing input boundaries.

32|5|Updated Mar 13, 2026

Search Tags:#threat-detection #ai-security #prompt-injection #output-filtering #input-sanitization

Ready to run

npx skills add https://github.com/jiayaoqijia/Ottie --skill prompt-injection-guard-jiayaoqijia

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: prompt-injection-guard
Source: https://github.com/jiayaoqijia/Ottie/tree/main/workspace/skills/safety/prompt-injection-guard
Command: npx skills add https://github.com/jiayaoqijia/Ottie --skill prompt-injection-guard-jiayaoqijia

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prompt injection defenses protect AI systems from malicious prompts that override system instructions, leak sensitive data, or bypass safeguards, ensuring safer and more reliable interactions.

Core Features & Use Cases

Input boundary enforcement that clearly separates user data from system prompts to prevent instruction leakage.
Threat detection and blocking of high-risk patterns such as jailbreak attempts and prompt tampering, with alerts for review.
Output filtering and context separation to prevent leakage of secrets and ensure external data are analyzed but not executed.

Quick Start

Enable the prompt-injection-guard with default rules and run a safety check on incoming user input.

Frequently Asked Questions about prompt-injection-guard

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

FAQPage Schema

How do I prevent prompt injection attacks in my AI chatbot?▼

Prompt injection defenses block malicious prompts by enforcing strict input boundaries and isolating user data from system prompts. This approach prevents instruction leakage, stops jailbreak attempts, and ensures safer AI interactions by treating external data as untrusted content.

What is the best way to isolate user input from system prompts?▼

The best way to isolate user input is by applying context separation and input boundary enforcement. This ensures external data is analyzed but never executed, preventing prompt tampering and safeguarding system instructions from being overridden by malicious user submissions.

How do I detect and block jailbreak attempts in AI assistants?▼

Detecting jailbreak attempts requires implementing threat detection rules that identify high-risk patterns in user input. This Skill automatically flags prompt tampering and direct jailbreak scenarios, generating alerts for review while blocking the malicious input from reaching the core model.

Can I use this prompt injection guard across different chat platforms?▼

Yes, this prompt injection guard applies to AI assistants and chat interfaces across platforms. It requires no specific dependencies, allowing you to integrate input sanitization and output filtering into various environments handling user data and external prompts.

How do I stop sensitive data leakage from AI assistant responses?▼

To stop sensitive data leakage, you implement output filtering and context separation. This ensures secrets remain protected and external data inputs are analyzed for risks rather than executed, preventing unauthorized data extraction through manipulated prompts.