Prompt Security Browser Extension
iidnankcocecmgpcafggbgbmkbcldmno
Risk Score
4.19
Risk Level:
Medium
Recommendation:
🟡 MEDIUM RISK — review
Top Risks
- Privacy policy is SentinelOne's generic corporate notice: scope_extension=false with data_collection+third_party_sharing=true — scores maximum privacy risk.
- <all_urls> host permission + content scripts on every site gives broad page-content access; justified for a GenAI DLP tool but high-impact if compromised.
- No CSP (MV3 mitigates somewhat, but no explicit policy declared); 12 external AI/enterprise hosts contacted.
- identity.email permission exposes signed-in user PII; no extension-scoped disclosure of how this is handled.
- 200K installs amplify blast radius if extension is ever compromised or sold.
Evidence
- privacy_policy_generic store Policy is SentinelOne corporate notice; scope_extension=false, data_collection=true, third_party_sharing=true → Privacy pillar maxed at 10.
- verified_publisher store Verified publisher badge present; Reputation floor applied at 2.0.
- host_permissions_all_urls manifest <all_urls> + content_scripts on http://* and https://* — full page access justified for DLP category.
- external_hosts_12 crx 12 distinct AI/enterprise endpoints: OpenAI, Claude, Copilot, Gemini, Google Docs, M365, Outlook, Bing, etc.
- no_csp manifest content_security_policy is null; MV3 provides some default isolation but no explicit hardening.
- no_code_findings crx code_findings_raw empty, obfuscation_score=0.0, no CVEs — clean code surface.
- maintenance_current store Last updated June 7, 2026 — 0 months stale; actively maintained.
- no_bad_hosts api threat_intel bad_host_hits, affiliate_hits, monetization_hits all empty.
Permissions Breakdown
- tabs medium Can read tab URLs and metadata across browser sessions.
- storage low Local config/state persistence; limited standalone risk.
- identity low OAuth token flow; scoped to extension auth needs.
- identity.email medium Reads signed-in user email; PII exposure if exfiltrated.
- alarms low Scheduled background tasks; low direct risk.
- declarativeNetRequest medium Can block/modify network requests; powerful but non-reading.
- downloads medium Can intercept or initiate file downloads.
- <all_urls> (host) high Content scripts on every page; broadest possible reach.
Pillar Scores
Permissions5.50
Reputation2.00
Network5.50
Webstore2.50
Maintenance0.00
Privacy10.00
Code Quality0.00
CVE Exposure0.00
Bookkeeping
Rubric v3.6
Scored at 2026-06-15 14:19
Listing SHA
d7614db4f42b…
Force block
— not fired
Score recovered
no
Elapsed
22.2s