Top Risks

• Privacy policy fetch failed — treated as no policy; data handling completely unknown for 100K users.
• Brand impersonation: mentions YouTube, ChatGPT, Claude without verified ownership of those brands.
• Content scripts run on ChatGPT, Claude, Gemini, YouTube — can read sensitive AI conversations and video data.
• function_constructor (new Function) and innerHTML sinks present — DOM-XSS and code-injection risk.
• Uninstall URL hijack detected — onUninstall redirects to third-party destination.

Evidence

• privacy_policy_fetch_failed api privacy_policy_classification.fetched==false (fetch_error:HTTPError); scored as +10.0 Privacy.
• brand_impersonation store brand_mention.is_impersonation==true; mentions youtube/chatgpt/claude; not verified_publisher. +2.0 Reputation.
• uninstall_url_hijack crx uninstall_url_hijack==true; +3.0 Webstore (uninstall URL hijack rule).
• content_scripts_broad_ai_origins manifest Scripts on ChatGPT, Claude, Gemini, YouTube, Mistral — high-value data surfaces for an AI summary tool.
• code_function_constructor crx new Function() in get-current-media JS; +2.5 Code Quality (function_constructor signal).
• dom_sink_innerhtml crx Two innerHTML-from-variable sinks; CSP present but sinks active; +0.5 each (+1.0 total) Code Quality.
• developer_name_missing store developer_name is empty string; no 'Offered by' display name. +1.0 Reputation.
• featured_by_google store is_featured_by_google==true; -2.0 Reputation (Featured badge).

Permissions Breakdown

• storage low Local data persistence only; low standalone risk.
• system.display low Read display config; no user data exfil path.
• content_scripts on AI/YouTube origins medium Scripts run on ChatGPT, YouTube, Gemini, Claude — can read page content including conversations.

Pillar Scores

Bookkeeping