Sign in with the account that started this chat to continue it.
Review https://grateful-minnow-9.convex.site. Inspect a sample flagged message and its reasoning without connecting mail or triggering alerts.
I’ll inspect Sentinel’s public demo only, using its sample data and avoiding setup, mail connection, forwarding, and any alert-triggering actions.
The demo exposes a read-only flagged case with both the forwarded email and Sentinel’s evidence panel. I’m checking the cited CRA guidance separately, then I’ll return to the case; I won’t use the dashboard’s “check an email,” handling, false-alarm, or setup controls.
## Review complete I inspected the public Sentinel demo at [grateful-minnow-9.convex.site/s/demo-family](https://grateful-minnow-9.convex.site/s/demo-family) in read-only mode. No mail account was connected, no email was forwarded, and I did not use “Check an email,” “mark handled,” “False alarm,” setup, or any alert-triggering control. ### Sample flagged message The selected pre-populated case was a supposed Canada Revenue Agency email: - Sender domain: `cra-secure-verify.com`, not an official CRA domain - Demanded prepaid gift-card payment within 24 hours - Used an HTTP link to the same unofficial domain - Threatened legal action and a warrant I did not visit the suspicious link. ### Sentinel’s reasoning Sentinel labeled it **Scam** with a **100/100 risk score**. Its explanation identified: 1. The sender-domain mismatch 2. The gift-card payment demand 3. The HTTP and unofficial destination 4. The CRA’s own scam guidance It also displayed a “What we checked” section, urgency indicators, the original forwarded message, and a plain-language reply advising Margaret not to reply, click, or pay. I independently checked the cited [CRA scam guidance](https://www.canada.ca/en/revenue-agency/corporate/scams-fraud/recognize-scam.html). It confirms that the CRA will not demand immediate payment through gift cards, ask for personal or financial information by email, or recommend following suspicious email links. The guardian alerts shown in the case were existing demo data; no alert was initiated during this review. The live mail-forwarding and alert workflow remains untested because the request was limited to safe, read-only inspection.