Top Risks

• Privacy policy is Google's generic policy — not scoped to this extension; admits data collection and 3rd-party sharing.
• script_src_dynamic in multiple bundles: dynamic script injection across all_urls is a high-impact XSS vector.
• new Function() with DOM attribute input in mmlWorker/bundle.js — potential arbitrary code execution from page content.
• scripting + <all_urls>: broad programmatic script injection on every site the user visits.
• identity.email permission collects user PII (email); combined with broad host access raises exfil concern.

Evidence

• privacy_policy_generic store Privacy URL points to Google's own account policy (scope_extension=false, data_collection=true, third_party_sharing=true) — not extension-specific.
• script_src_dynamic_multiple_files crx Dynamic <script> element creation found in calculator/bundle.js, mathDiscoverGSuiteBundle.js, mathDiscoverabilityBundle.js, formBundle.js, mathjaxFrame/bundle.js.
• function_constructor_dom_attr crx new Function('item', e.getAttribute('jsTest')) in mmlWorker/bundle.js — executes DOM attribute as code.
• dom_innerHTML_sink crx innerHTML from variable in browserAction/bundle.js, htmlEditorApi.js, content/6067.js, mathjaxFrame/bundle.js — DOM-XSS sinks.
• broad_host_scripting manifest host_permissions=<all_urls> + scripting permission + content_scripts on <all_urls>; justified for Accessibility but high capability.
• is_featured_by_google store Google Featured badge present; reduces reputation risk but does not override code/privacy concerns.
• cve_findings_empty crx No CVEs detected in bundled libraries (jquery 3.5.1/3.6.0/3.7.1, react 16.13.1/17.0.2). CVE pillar = 0.
• sandbox_csp_unsafe_eval manifest sandbox CSP includes 'unsafe-eval' — eval permitted inside sandboxed frames, raises risk when combined with dynamic script loading.

Permissions Breakdown

• activeTab low Grants access to active tab on user action; scoped and low risk.
• tabs medium Can read tab URLs and metadata across sessions.
• alarms low Schedules background tasks; no data access.
• storage low Local/sync storage; no exfil risk alone.
• identity medium OAuth token access; can retrieve user identity tokens.
• identity.email medium Explicitly requests user email via OAuth — PII exposure.
• gcm low Push messaging; low direct risk but enables remote triggering.
• scripting high Programmatic script injection into pages; HIGH risk with <all_urls>.
• <all_urls> (host_permissions) high Broad host access across all sites; paired with scripting amplifies risk.

Pillar Scores

Bookkeeping