When a shopping agent parses your page it reads user content verbatim. Attackers hide instructions inside it to hijack the agent, this flags the patterns before they reach an agent's context.
Heuristic scan for instruction-shaped patterns, hidden HTML comments, and redirect URLs. Mitigate by stripping comments and control sequences at ingestion, and serving agent-facing content from structured fields only.
Why this matters
Agent-targeted prompt injection means attackers plant instructions inside public product descriptions, reviews, or HTML metadata. When a shopping agent reads the page, the injected text can override its logic, pushing it to redirect a purchase, leak a token, or recommend an inferior product. Your user-generated content is now an executable surface.
This scanner flags the tell-tale patterns. It runs entirely in your browser and stores nothing.
How to use it
- Paste a review or user-generated text.
- See flagged injection patterns, hidden comments and obfuscated URLs.
- Sanitise the content before it renders on your page.
Frequently asked questions
What is prompt injection in user content?
Instruction-shaped text hidden in reviews or comments, HTML comments, 'ignore previous instructions', redirect payloads, that tries to hijack an AI agent parsing your page. Rendered on a public product page, it becomes an executable attack surface.
How do I protect against it?
Sanitise user-generated content at ingestion: strip HTML comments and control sequences, and flag instruction-shaped patterns before storage. The scanner shows which reviews are clean, suspicious or malicious.
Is the UGC Prompt-Injection Scanner free to use?
Yes. The UGC Prompt-Injection Scanner is completely free, with no sign-up, no usage limits and no watermark on the output.
Is my data private?
Yes. The UGC Prompt-Injection Scanner runs entirely in your browser. Nothing you paste or enter is uploaded, stored or sent to any server.
Related tools and reading
Keep going with the playbooks and tools behind this one.