<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom"><title>Todd Schiller - dark patterns</title><link href="https://toddschiller.com/" rel="alternate"></link><link href="https://toddschiller.com/feeds/tag/dark-patterns.atom.xml" rel="self"></link><id>https://toddschiller.com/</id><updated>2026-06-03T00:00:00-04:00</updated><subtitle>Human ✘ Artificial Intelligence</subtitle><entry><title>agent-browser-shield June 3 update: 14 new rules and the Chrome Web Store listing</title><link href="https://toddschiller.com/blog/agent-browser-shield-june-03-update.html" rel="alternate"></link><published>2026-06-03T00:00:00-04:00</published><updated>2026-06-03T00:00:00-04:00</updated><author><name>Todd Schiller</name></author><id>tag:toddschiller.com,2026-06-03:/blog/agent-browser-shield-june-03-update.html</id><summary type="html">14 new rules shipped in agent-browser-shield, now installable from the Chrome Web Store, and using it for daily driving.</summary><content type="html">&lt;p&gt;Two days after the alpha announcement, agent-browser-shield has a Chrome Web
Store listing and 14 new protection rules.&lt;/p&gt;
&lt;h2&gt;Install from the Chrome Web Store&lt;/h2&gt;
&lt;p&gt;The extension is live at
&lt;a href="https://chromewebstore.google.com/detail/agent-browser-shield/gnejacdioaelglahihpagpfjpddpnamd"&gt;chromewebstore.google.com/detail/agent-browser-shield&lt;/a&gt;.
One click instead of unpacked-from-source. The prebuilt ZIP and source-build
paths stay for Browserbase and other runtimes that need an unpacked
extension.&lt;/p&gt;
&lt;h2&gt;New rules: handling prompt injection and context pollution in invisible surfaces&lt;/h2&gt;
&lt;p&gt;A browser-use agent reads surfaces a sighted user never looks at. The new
rules close them:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code&gt;&amp;lt;noscript&amp;gt;&lt;/code&gt; blocks (never rendered with JS on, but agents walk them)&lt;/li&gt;
&lt;li&gt;Poisoned &lt;code&gt;&amp;lt;meta&amp;gt;&lt;/code&gt; description and &lt;code&gt;&amp;lt;title&amp;gt;&lt;/code&gt; (the compact &amp;quot;what is this
page&amp;quot; answer many agents pull first)&lt;/li&gt;
&lt;li&gt;JSON-LD &lt;code&gt;&amp;lt;script&amp;gt;&lt;/code&gt; blocks (cited as the &amp;quot;trusted summary&amp;quot; of a page)&lt;/li&gt;
&lt;li&gt;&lt;code&gt;aria-label&lt;/code&gt;, &lt;code&gt;alt&lt;/code&gt;, &lt;code&gt;title&lt;/code&gt;, &lt;code&gt;placeholder&lt;/code&gt;, and SVG &lt;code&gt;&amp;lt;title&amp;gt;&lt;/code&gt; / &lt;code&gt;&amp;lt;desc&amp;gt;&lt;/code&gt;
/ &lt;code&gt;&amp;lt;text&amp;gt;&lt;/code&gt; (a11y-tree carriers)&lt;/li&gt;
&lt;li&gt;Unicode tag characters, bidi overrides, and zero-width payloads&lt;/li&gt;
&lt;li&gt;Long base64 / hex / percent-encoded blobs (the &amp;quot;decode this and follow
it&amp;quot; pattern)&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;New rules: trust laundering&lt;/h2&gt;
&lt;p&gt;&lt;code&gt;link-spoof-annotate&lt;/code&gt; flags Cyrillic homoglyphs and anchors whose visible
text doesn't match the href apex. &lt;code&gt;disguised-ad-flag&lt;/code&gt; collapses native
advertorials (Sponsored / Paid Post) that share DOM shape with editorial.
&lt;code&gt;trust-badge-annotate&lt;/code&gt; and &lt;code&gt;schema-trust-sanitize&lt;/code&gt; ship off by default
while we assess their false-positive rates.&lt;/p&gt;
&lt;h2&gt;The daily-driver surprise&lt;/h2&gt;
&lt;p&gt;I've started running it on my own daily-driver browser, not just agent runs.
There have been some funny quirks (e.g., flagging GitHub issue links with &amp;quot;.md&amp;quot;
in the link text as suspicious and hiding the GitHub issue template modal).
However, overall it's been a positive on my browsing experience. So, we'll be
experimenting with making more annotations visible to humans and
multi-modal LLMs.&lt;/p&gt;
&lt;p&gt;⭐ &lt;a href="https://github.com/pixiebrix/agent-browser-shield"&gt;https://github.com/pixiebrix/agent-browser-shield&lt;/a&gt;&lt;/p&gt;
</content><category term="AI"></category><category term="AI"></category><category term="agents"></category><category term="browser security"></category><category term="prompt injection"></category><category term="dark patterns"></category><category term="PixieBrix"></category></entry></feed>