Content script

#How a content script works

An extension declares content scripts in its manifest, with match patterns for the sites they apply to. An extension can also inject a script at runtime with the scripting API. When a matching page loads, the browser runs the script in that page.

The script shares the page's DOM. It has its own JavaScript variables, so it cannot read variables that the page defines. To talk to the rest of the extension, it sends messages, usually to the background service worker.

#Why it matters for AI agents

Content scripts let an extension see the page as the user sees it, including content that loads after the page starts. That makes them useful for reading results or filling in forms.

The same access carries a risk. The script runs inside every page it matches. Keep match rules narrow, and ask only for the permissions the extension needs.

#In practice with AI agents

An agent-facing extension might use a content script to read the visible text of a results page and send it to the background worker. The page text was written by the site, so treat it as content, not as instructions for the agent.

#Related terms

• Manifest V3
• Extension service worker
• Same-origin policy
• Native messaging

#FAQ

#Can a content script read the page's JavaScript variables?

No. A content script runs in an isolated world. It shares the DOM with the page, but not the page's JavaScript objects or variables.

#How is a content script different from a service worker?

A content script runs inside a web page, with access to that page's DOM. An extension service worker is the extension's background script. It has no page DOM, starts on events and runs apart from any one page.