skills.rest
Kaleb-Rupe
Kaleb-RupeCOMMUNITY
Explore all skills by this author →
𝕏@MightieMagsGitHub

PromptInjection

Tests LLM applications for prompt injection vulnerabilities using structured reconnaissance and attack taxonomies.

Updated Jul 10, 2026
Search Tags:#owasp#penetration testing#jailbreak#prompt injection#llm security#red teaming#rag poisoning
One-click install
npx skills add https://github.com/Kaleb-Rupe/aurora --skill promptinjection-kaleb-rupe
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: PromptInjection
Source: https://github.com/Kaleb-Rupe/aurora/tree/main/claude/skills/Security/PromptInjection
Command: npx skills add https://github.com/Kaleb-Rupe/aurora --skill promptinjection-kaleb-rupe

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Security teams lack a systematic way to assess LLM-powered applications against prompt injection, jailbreaking, RAG poisoning, and system prompt extraction attacks during authorized penetration tests. ## Core Features & Use Cases - Comprehensive Attack Taxonomy: Covers 15 attack categories including direct injection, indirect injection, multi-stage chains, RAG poisoning, token manipulation, and obfuscation techniques. - 7-Phase Reconnaissance Methodology: Browser-automation-driven discovery of DOM inputs, JavaScript endpoints, API parameters, and AI components to map the attack surface. - Defense Evaluation & Tooling Guidance: Documents defense-in-depth layers (sanitization, instruction/data separation, output filtering) and compares automated scanners like Promptfoo, Garak, and PyRIT. - Use Case: During an authorized engagement against a customer support chatbot, run the reconnaissance workflow to enumerate the /api/chat endpoint and its parameters, then execute prioritized injection payloads and document findings with remediation guidance. ## Quick Start Ask the assistant to test an authorized target application for prompt injection vulnerabilities following the reconnaissance and attack taxonomy methodology.

Frequently Asked Questions about PromptInjection

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

FAQPage Schema
How do I test an LLM application for prompt injection vulnerabilities?▼

Start with the 7-phase reconnaissance methodology to map inputs, API endpoints, and AI components using browser automation. Then execute prioritized attack payloads from the taxonomy, beginning with direct injection and system prompt extraction tests, and document all findings.

What automated tools test for prompt injection attacks?▼

Promptfoo suits CI/CD integration with AI-generated attacks, Garak provides a research-backed static probe library for baseline scans, and PyRIT offers Python-scriptable enterprise testing with Azure integration. Combining tools with manual testing gives the best coverage.

Is written authorization required before prompt injection testing?▼

Yes, written authorization is mandatory before any testing. The methodology requires documenting the target, scope, authorized contacts, and testing window in an AUTHORIZATION.md file, and testing must stop immediately if authorization is unclear.

How effective are prompt injection defenses like input sanitization?▼

No single defense is sufficient; documented bypass rates exceed 90% for most individual controls. Effective protection requires defense-in-depth combining input sanitization, instruction/data separation, output filtering, privilege isolation, and RAG-specific defenses.

What is indirect prompt injection and why is it dangerous?▼

Indirect prompt injection embeds malicious instructions in external content like web pages, documents, emails, or RAG knowledge bases that the LLM later processes. It is stealthy, persistent, and can affect many users from a single poisoned source, with documented success rates up to 90%.

Related Skills

Explore More Software Engineering Skills→
AnandaAnugrahHandyantoAnandaAnugrahHandyanto

web-pentest

Automate authorized web application penetration testing with phase-based workflows and evidence capture.

Community
Advanced
commercetoolscommercetools
154

security-auditor

Audit code repositories for vulnerabilities and generate remediation reports.

Official
Advanced
brucesongsbrucesongs
60

llm-red-team

Expose LLM system vulnerabilities via a six-phase red-team workflow.

Community
Advanced
Q16GQ16G
83

graybox-p0

Orchestrate gray-box testing by combining white-box candidate discovery with black-box request validation.

Community
Intermediate