PromptInjection

You can test for prompt injection vulnerabilities by running the CompleteAssessment workflow, which executes end-to-end security testing. It identifies jailbreaks, system prompt leakage, context manipulation, and guardrail bypasses within authorized engagements.

Direct prompt injection testing targets immediate jailbreaks and system prompt extraction, while indirect injection testing focuses on RAG poisoning and context manipulation. The skill applies 9-step testing sequences to validate defenses against both multi-stage attack vectors.

Yes, you can simulate RAG poisoning attacks to validate LLM defenses. The skill includes RAG poisoning simulations and multi-stage attack capabilities as part of its workflow-driven reconnaissance and direct or indirect injection testing sequences.

Yes, you need written authorization to perform prompt injection testing. The skill requires written authorization, artifact collection, and configurable workflows to execute security testing safely and ensure compliant vulnerability assessment reporting.

The reconnaissance phase of LLM security testing includes a 7-phase reconnaissance workflow designed to gather artifacts and identify potential attack surfaces. This phase precedes the 9-step direct and indirect injection testing sequences.