Unverified60% confidenceFactExact time
Common techniques for concealing prompt injections include white text, CSS hidden elements, HTML comments, and Unicode special characters.
2
Sources
60%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
Prismata Analysis: How to Defend Web Agents Against Cross-Site Prompt Injection Attacks
hackernewshackernews7/10/2026
Related Claims
UnverifiedCSS视觉渲染与DOM实际内容之间存在分离特性，字符可通过白底白字、font-size:0或opacity:0在视觉上不可见但仍存在于HTML文档中，复制粘贴会带走68% similarUnverified间接注入攻击手法包括在网页嵌入与背景色相同的隐藏文字、在PDF元数据或注释字段植入越狱指令、利用Markdown语法伪装恶意提示67% similarUnverifiedZero-visibility styling refers to exploiting differences between the Markdown, HTML, or PDF rendering layer and the model's parsing layer to hide content, such as CSS color:white, font-size:0, display:none, PDF invisible text layers, metadata fields, and HTML comments, which are retained during text extraction and enter the model context.67% similarUnverifiedFor interactive components such as tabs, accordions and "read more", the text should be kept in the DOM and hidden with CSS rather than loaded on click, so that it is visible to AI crawlers.66% similarUnverifiedIn resume screening scenarios, an attacker can embed HTML content invisible to the human eye in a PDF (such as CSS-hidden styles, metadata, and HTML comments), and this content is retained when the model reads the plain text and may be executed as instructions.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/487958API
curl https://kongchang.com/api/v1/knowledge/claims/487958MCP
get_claim(id=487958)