ai_summarizer_context

CRITICAL: The Chrome extension component (`extension/src/sidepanel.js`) directly parses JSON-like strings from the LLM's streaming output and executes them as browser automation commands via `chrome.tabs.sendMessage` to `content_script.js` when `automation_enabled` is true. This creates a direct pipeline for LLM-generated commands to control the browser, allowing arbitrary actions (e.g., clicking links, typing, reading page content). This is a significant security risk, as a compromised or 'misaligned' local LLM could lead to data exfiltration, phishing, or other malicious activities without explicit user confirmation for each command. While the desktop app's MCP tool calls appear to be user-initiated via a UI panel, the browser extension's automation is direct.