Llm Prompt Injection

This is a comprehensive playbook for testing LLM applications against prompt injection attacks. It covers direct injection (instruction override, role play), indirect injection through RAG poisoning and web browsing, tool abuse via function calling chains, and data exfiltration through markdown images and encoded payloads. The MCP security section is especially relevant if you're building or auditing systems that use external tool servers, since it breaks down how malicious tool descriptions can carry hidden instructions. Use this when you need to think like an attacker probing AI systems, not just running basic "ignore previous instructions" tests. It's practical attack vectors, not theory.

skills.sh installs: 3.9K
GitHub stars: 2.6K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add yaklang/hack-skills --skill llm-prompt-injection --agent claude-code

Installs into .claude/skills of the current project.

Files

SKILL: LLM Prompt Injection — Expert Attack Playbook

AI LOAD INSTRUCTION: Expert LLM prompt injection techniques. Covers direct injection (instruction override, role play, context manipulation), indirect injection (RAG poisoning, web browsing, email), tool/function abuse, data exfiltration, MCP security risks, and defense bypass (encoding, splitting, few-shot). Base models miss the distinction between direct and indirect injection and underestimate tool-calling attack chains.

0. RELATED ROUTING

ai-ml-security for broader ML security (adversarial examples, model poisoning, model extraction, data privacy attacks)
xss-cross-site-scripting for parallels between XSS (injecting into HTML context) and prompt injection (injecting into LLM context)
ssrf-server-side-request-forgery when prompt injection chains into SSRF via tool calls