Please report security issues privately through GitHub Security Advisories rather than as a public issue.

Expect an acknowledgement within a few days. This is a solo project, so response times are best effort rather than contractual.

This extension handles API keys and page content, so the areas most worth scrutiny are:

• API key handling. Keys live in chrome.storage.local and are read only by the service worker. Content scripts never receive one, and no key is sent anywhere except the provider that issued it. chrome.storage.local is not encrypted — that is a documented limitation, not a vulnerability report.
• Rendering model output. src/shared/ui/markdown.js escapes HTML before applying any markup, because page text reaches the model and a hostile page can attempt prompt injection to get markup into the response. The summary modal renders inside the page's document, so a bypass there would execute in page context. This is the highest-value area to look at.
• Injected content scripts. The extension declares no content_scripts and injects on demand under activeTab, so it should have no ability to read pages the user has not acted on. A path to reading arbitrary pages would be a genuine finding.
• Web accessible resources. Feature modules are listed in web_accessible_resources so the content entry can import them. They contain no secrets, but confused-deputy issues are in scope.

• The unencrypted-at-rest nature of chrome.storage.local, which Chrome does not offer an alternative to
• Content sent to a third-party AI provider that the user explicitly configured
• Vulnerabilities in the AI providers themselves