Sublime Security

Giving a security agent the right amount of control

I lead design for ADÉ, Sublime’s agent that writes detection rules to address missed threats.

When
2025 to now
Role
Staff Product Designer
Team
1 PM, engineers, 2 designers
Outcome
New rule in under 2 hours, not days

Action required: payroll profile verification

Completed in 48mJob ID: 019f3d5a-ff42-7706-b615-a83297b00479

ADÉ Recommended Rule

99% match in a 14-day hunt, with 40 malicious matches and 2 possible false positives.

AcceptReject
Rule NameCredential phishing: payroll QR code lure with newly registered domain
SeverityHigh Severity
DescriptionDetects inbound payroll-themed messages that contain QR-code links to newly registered credential-harvesting domains.
LabelsCredential PhishingQR codeSocial engineering
ActionQuarantine
Rule MQL
1type.inbound2and any(attachments,3 .file_extension in~ ("png", "jpg", "jpeg", "pdf")4 and any(file.explode(.),5 .scan.qr.type == "url"6 and ml.link_analysis(.scan.qr.data).credphish.disposition == "phishing"7 )8)9and any(ml.nlu_classifier(body.current_thread.text).intents,10 .name in ("payroll", "credential_theft")11 and .confidence in ("medium", "high")12)13and network.whois(sender.email.domain).days_old <= 45

ADÉ hunted across 14 days and found 0 message groups (0 total messages)

Recommended Rule is a 0% match with 0 malicious matches and may introduce 0 false positives (Benign).
0 Malicious0 Spam0 Benign

ADÉ worked for 48m

Collaborated with ASA Agent to generate a threat reportValidated recommendation with 1 hunt

Patrick Morgan triggered ADÉ at Jul 3, 2026, 2:12 PM

SubjectAction required: payroll profile verificationSenderpayroll-verify@employee-benefits.coRecipientalex.morgan@acme.comASA VerdictMalicious
ADÉ writes a rule your team can read

Each missed threat needed a new rule

An engineer wrote and tested each rule by hand, so a fix took days or weeks.

Action required: payroll profile verification

FromPayroll Operations <payroll-verify@employee-benefits.co>Toalex.morgan@acme.com
ASA Verdict

Malicious

Executive Summary

This message impersonates Payroll Operations and urges the recipient to verify their payroll profile through an unfamiliar domain. The sender fails DMARC, does not align with the organization’s trusted HR infrastructure, and uses credential-harvesting language under the pretext of an upcoming pay cycle.

Matched Detection Rules (0)

No Detection Rules matched this message group.

Message Content

User ViewText ViewHTML ViewEML View

CAUTION: This message originated outside of your organization.

From: Payroll Operations <payroll-verify@employee-benefits.co>

Action required

Your payroll profile requires verification before the next pay cycle. Review and confirm your direct deposit details to avoid a delay in payment.

Review Status

Review Status⚠ Unreviewed
Select Classification
Malicious
Spam
Graymail
Benign

Other

Send EML to Rule Editor
Share feedback
ASA marks the message malicious, and no rule caught it

The agent writes rules that people can read and change

Other vendors update a central model. ADÉ writes a rule each team can read, edit and turn off.

Action required: payroll profile verification

Completed in 48mJob ID: 019f3d5a-ff42-7706-b615-a83297b00479

ADÉ Recommended Rule

99% match in a 14-day hunt, with 40 malicious matches and 2 possible false positives.

AcceptReject
Rule NameCredential phishing: payroll QR code lure with newly registered domain
SeverityHigh Severity
DescriptionDetects inbound payroll-themed messages that contain QR-code links to newly registered credential-harvesting domains.
LabelsCredential PhishingQR codeSocial engineering
ActionQuarantine
Rule MQL
1type.inbound2and any(attachments,3 .file_extension in~ ("png", "jpg", "jpeg", "pdf")4 and any(file.explode(.),5 .scan.qr.type == "url"6 and ml.link_analysis(.scan.qr.data).credphish.disposition == "phishing"7 )8)9and any(ml.nlu_classifier(body.current_thread.text).intents,10 .name in ("payroll", "credential_theft")11 and .confidence in ("medium", "high")12)13and network.whois(sender.email.domain).days_old <= 45

ADÉ hunted across 14 days and found 0 message groups (0 total messages)

Recommended Rule is a 0% match with 0 malicious matches and may introduce 0 false positives (Benign).
0 Malicious0 Spam0 Benign

ADÉ worked for 48m

Collaborated with ASA Agent to generate a threat reportValidated recommendation with 1 hunt

Patrick Morgan triggered ADÉ at Jul 3, 2026, 2:12 PM

SubjectAction required: payroll profile verificationSenderpayroll-verify@employee-benefits.coRecipientalex.morgan@acme.comASA VerdictMalicious

Accept Recommendation

This will create the recommended detection rule and apply an Auto-Quarantine action. You’ll be able to edit the Rule at any time after it’s created.

Leave a comment (optional)
Comment
CancelAccept & Activate Rule
ADÉ recommends a rule, with the hunt that tested it

The agent starts where the team finds the threat

One click beside the verdict starts a job, and the job page shows each step ADÉ does.

Action required: payroll profile verification

FromPayroll Operations <payroll-verify@employee-benefits.co>Toalex.morgan@acme.com
ASA Verdict

Malicious

Executive Summary

This message impersonates Payroll Operations and urges the recipient to verify their payroll profile through an unfamiliar domain. The sender fails DMARC, does not align with the organization’s trusted HR infrastructure, and uses credential-harvesting language under the pretext of an upcoming pay cycle.

Matched Detection Rules (0)

No Detection Rules matched this message group.

Message Content

User ViewText ViewHTML ViewEML View

CAUTION: This message originated outside of your organization.

From: Payroll Operations <payroll-verify@employee-benefits.co>

Action required

Your payroll profile requires verification before the next pay cycle. Review and confirm your direct deposit details to avoid a delay in payment.

Review Status

Review Status⚠ Unreviewed
Select Classification
Malicious
Spam
Graymail
Benign

Other

Send EML to Rule Editor
Share feedback

ADÉ (Autonomous Detection Engineer)

ADÉ recommends new Detection Rules for malicious, spam, or graymail messages that do not match any Detection Rules.One click starts ADÉ
ADÉ waits on the same page as the verdict

Each rule shows what it would catch before it goes live

ADÉ runs each rule on the last 14 days of email and shows what it catches and wrongly blocks.

Action required: payroll profile verification

Completed in 48mJob ID: 019f3d5a-ff42-7706-b615-a83297b00479

ADÉ Recommended Rule

99% match in a 14-day hunt, with 40 malicious matches and 2 possible false positives.

AcceptReject
Rule NameCredential phishing: payroll QR code lure with newly registered domain
SeverityHigh Severity
DescriptionDetects inbound payroll-themed messages that contain QR-code links to newly registered credential-harvesting domains.
LabelsCredential PhishingQR codeSocial engineering
ActionQuarantine
Rule MQL
1type.inbound2and any(attachments,3 .file_extension in~ ("png", "jpg", "jpeg", "pdf")4 and any(file.explode(.),5 .scan.qr.type == "url"6 and ml.link_analysis(.scan.qr.data).credphish.disposition == "phishing"7 )8)9and any(ml.nlu_classifier(body.current_thread.text).intents,10 .name in ("payroll", "credential_theft")11 and .confidence in ("medium", "high")12)13and network.whois(sender.email.domain).days_old <= 45

ADÉ hunted across 14 days and found 0 message groups (0 total messages)

Recommended Rule is a 0% match with 0 malicious matches and may introduce 0 false positives (Benign).
0 Malicious0 Spam0 Benign

ADÉ worked for 48m

Collaborated with ASA Agent to generate a threat reportValidated recommendation with 1 hunt

Patrick Morgan triggered ADÉ at Jul 3, 2026, 2:12 PM

SubjectAction required: payroll profile verificationSenderpayroll-verify@employee-benefits.coRecipientalex.morgan@acme.comASA VerdictMalicious
With automation on, the rule arrives with its hunt result

Each team decides how much the agent does alone

A team sets when ADÉ starts a job and the test score a rule needs to go live.

ADÉ Automation

Automatically run ADÉ and accept generated rules when validation meets your threshold.

AuthorSublime Security

StatusInactive

Configure ADÉ Jobs

Choose, by verdict, when ADÉ starts on its own and when it accepts the rule it writes.

VerdictAuto-start ADÉ jobAlertsRule auto-acceptance threshold
MaliciousNo Alert
Send Slack alert
Send webhook
Require review
Recommended (99%)
Custom
Require review
SpamNo AlertRequire review
GraymailNo AlertRequire review
At first, ADÉ runs only when someone starts it

Missed threats now get a rule in under 2 hours

A fix that took an engineer days or weeks now arrives as a rule ready to accept.

ADÉ (Autonomous Detection Engineer)

Creates detections tailored to your environment.

Go to Settings

Performance (Last 30 days)

0.0 hrs

Time to Coverage

0

Hours Saved

0.0k

Attacks Blocked

Jobs

All recommendation statuses
Message SubjectASA VerdictStarted ByStartedReport StatusRecommendation Status
Invoice payment confirmation required todayMaliciousADÉ9h agoIn Progress—
Benefits enrollment form from rare senderGraymailADÉ18h agoCompletedAuto-Accepted
DocuSign: updated employee handbook signature requestSpamAJ Williams1d agoCompletedAccepted
Action required: payroll profile verificationMaliciousADÉ1d agoCompletedNeeds Review
Calendar invite with external benefits linkSpamAJ Williams1d agoCompletedRejected
Urgent wire transfer details for reviewMaliciousADÉ2d agoCompletedAuto-Accepted
Shared voicemail transcript is readyGraymailFatima Al-Sayed12d agoCompletedAccepted
Payroll QR credential harvest reportMaliciousADÉ15d agoCompletedAuto-Accepted
ADÉ’s page: the time to cover a missed threat, and the jobs behind it