Prompt Injection Audit · itachi3355 bundle Audit an AI system you own or are authorized to test for prompt-injection and tool-poisoning vulnerabilities, then produce a prioritized findings report with concrete mitigations. Covers system prompts, tool/function descriptions, RAG and retrieval pipelines, MCP server integrations, and multi-turn agent flows. Use this skill whenever the user wants to security-review, red-team, threat-model, pen-test, or "check for injection" on an LLM app, agent, chatbot, RAG system, or MCP setup — or asks "is my agent vulnerable to prompt injection?", "audit my system prompt", "review my tool definitions for tool poisoning", or "harden my AI pipeline". Frame every engagement as authorized defensive testing of the user's own system.
