PromptInjection

Identify prompt injection vulnerabilities in LLM-powered applications through structured testing workflows.

Updated Aug 23, 2026

Search Tags:#defense-in-depth #reconnaissance #workflows #pentest #prompt-injection #llm-security

Ready to run

npx skills add https://github.com/larsboes/pai-marketplace --skill promptinjection-larsboes

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: PromptInjection
Source: https://github.com/larsboes/pai-marketplace/tree/main/marketplace/plugins/security/skills/PromptInjection
Command: npx skills add https://github.com/larsboes/pai-marketplace --skill promptinjection-larsboes

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

PromptInjection provides a structured framework to test and validate LLM security against prompt injection, jailbreaks, and guardrail bypasses in authorized engagements.

Core Features & Use Cases

End-to-end testing workflows (Reconnaissance, Direct Injection, Indirect Injection, Multi-Stage Attacks)
Integrations with testing tools (Promptfoo, Garak, PyRIT) and workflow automation
Defense-first methodology focused on containment, logging, and responsible disclosure
Use cases include security assessments of chatbots, enterprise AI apps, and compliance-driven reviews

Quick Start

Run the CompleteAssessment workflow to perform an end-to-end prompt injection security assessment on a target AI application.

Frequently Asked Questions about PromptInjection

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

FAQPage Schema

How do I test LLM applications for prompt injection vulnerabilities?▼

To test LLM applications for prompt injection vulnerabilities, you can run an end-to-end security assessment workflow that covers reconnaissance, direct and indirect injections, and RAG poisoning. This structured testing identifies jailbreaks and guardrail bypasses in authorized deployments.

What is the best way to perform a pentest on enterprise AI chatbots?▼

The best way to pentest enterprise AI chatbots is using a defense-in-depth methodology that applies browser automation for direct and indirect prompt injection testing. This approach validates LLM security while emphasizing containment, logging, and responsible disclosure.

Can I use Promptfoo and Garak for multi-stage LLM attack testing?▼

Yes, you can use Promptfoo, Garak, and PyRIT for multi-stage LLM attack testing through structured workflow automation. These testing tool integrations help security teams systematically demonstrate complex prompt injection vulnerabilities across targeted client deployments.

Do I need written authorization to run prompt injection security assessments?▼

Yes, you need explicit written authorization before running prompt injection security assessments. Authorized testing ensures compliance and validates that your reconnaissance, RAG poisoning, and multi-stage attack workflows follow responsible disclosure practices.

How does indirect prompt injection testing work for RAG poisoning?▼

Indirect prompt injection testing for RAG poisoning works by simulating malicious inputs embedded in external data sources retrieved by the LLM. Structured workflows identify these vulnerabilities by automating browser interactions and analyzing defense containment mechanisms.

What are the limitations of automated LLM security testing workflows?▼

Limitations of automated LLM security testing workflows include the strict requirement for written authorization and the need for defense analysis to validate findings. Testing focuses on demonstrating vulnerabilities rather than providing automated remediation for prompt injection flaws.