← Back to Labs

AI Context Poisoning & Prompt Injection

Step through indirect prompt injection in RAG document ingestion pipelines, adversarial embedding vectors, and XML context isolation boundaries

1. INGESTIONCLEAN PDFClean PDF ChunksSTATUSCLEAN INGESTEnterprise_PDF.pdf2. VECTOR DB1536-D Cosine SpaceClean #0Query qNEAREST NEIGHBORTOP-1: CLEANCosine Sim = 0.623. PROMPTSystem PromptClean ContextUser QuestionISOLATION BARRIERUNPROTECTEDRaw Token Concatenation4. LLM & GUARDLLMTransformer CoreReady for inferenceINFERENCE CONTROLIDLE / PENDINGStandard Pipeline
STEP 1 OF 6

1. Clean Document Vector Indexing

In a standard Retrieval-Augmented Generation (RAG) system, trusted corporate PDF documents are parsed into text chunks, converted into vector embeddings, and indexed in a vector database for quick semantic search.

Arrow keys to navigate · R to reset

Tap dots to jump to any step

Read the full article →Take the quiz →