Top Risks

• Critical CVE-2021-23358 in bundled underscore@1.8.3 enables arbitrary code execution; not patched to 1.12.1+.
• Privacy policy fetched but scope_extension=false and admits data_collection+third_party_sharing — D rule triggers +10.0.
• brand_mention.is_impersonation=true for LinkedIn without verified publisher — +2.0 reputation hit.
• new Function() constructor found in guest and background scripts — code quality risk with no CSP present.
• MV3 with no CSP: underscore CVE (DOM-manipulation lib) triggers ×1.5 CVE amplifier, pushing CVE pillar to 7.0.

Evidence

• critical_cve_bundled_lib crx underscore@1.8.3 bundles CVE-2021-23358 (critical ACE); fixed_in 1.12.1 — library not upgraded.
• high_cve_bundled_lib crx underscore@1.8.3 bundles CVE-2026-27601 (high DoS via recursion); fixed_in 1.13.8.
• no_csp manifest content_security_policy is null; no CSP defined — ×1.5 CVE amplifier applies.
• privacy_policy_generic_admits_sharing store Policy fetched but scope_extension=false, data_collection=true, third_party_sharing=true → Privacy pillar +10.0.
• brand_impersonation store brand_mention.is_impersonation=true for LinkedIn; developer not verified_publisher nor featured by Google on impersonation check.
• function_constructor crx new Function() constructor present in guest_script.js and background_script.js — dynamic code construction risk.
• is_featured_by_google store Extension carries Google Featured badge — partial trust signal, applied -2.0 reputation discount.
• host_permission_linkedin_only manifest Host access scoped to *.linkedin.com only; justified for LinkedIn content tool — no broad host access.

CVE Exposures (2)

Permissions Breakdown

• storage low Local key-value store; no cross-site or exfil risk on its own.
• *://*.linkedin.com/* medium Host permission scoped narrowly to linkedin.com; enables content-script injection on that domain.

Pillar Scores

Bookkeeping