Practical prompt-injection guides

Prompt Injection Check explains how to treat AI inputs as untrusted data, recognize common attack patterns, and design safer review steps.

Browse by goal

Learn: what prompt injection is, examples, indirect attacks, and jailbreak differences.

Build: secure RAG, agent security, and the prevention checklist.

Check: emails, PDFs, and scanner limitations.

A practical review checklist for forwarded messages, hidden instructions, and suspicious requests.

A plain-language explanation of direct attacks, authority confusion, and why ordinary text can become instructions.

How attacks hide in content retrieved by an AI and how to separate data from executable instructions.

Concrete controls for permissions, tool use, human approval, and safe handling of untrusted content.

Why deterministic warnings are useful, where they fail, and how to combine them with broader security controls.

Safe, redacted examples of common warning signs and how to review them.

Understand the difference and choose layered defenses.

Practical controls for permissions, tools, secrets, and approvals.

Keep retrieved documents untrusted while building useful RAG workflows.

Review PDF text and active content safely before AI use.

Quick answers about scanners, privacy, and safer AI workflows.