Skip to main content

Phishing triage

How an AI SOC triages reported and flagged phishing: work the campaign as one group, check links and attachments, find who clicked, and reply to the reporter.

Use case
Updated

The short answer

Phishing triage decides whether a reported or flagged email is malicious, who else received it, and who acted on it. In Cotool, a response agent gathers that evidence from email security, threat intel, and web proxy tools, then emails the employee who reported it.[1][2][3]

The problem

Phishing can be targeted at one person or sent as mass spam, and it carries malicious attachments or links, so one report may stand for many inboxes.[8]

Sublime groups similar messages under one canonical_id, so 100 users receiving the same email show up as a single group. Without that grouping, analysts work the same campaign over and over.[9]

User-reported phishing adds a reply loop: someone has to tell the reporter what happened to the email they flagged.[3]

How an AI SOC handles it

  1. Read the message group

    The agent opens the flagged message or Sublime message group and pulls Sublime's ActionMessage analysis into the alert.[4][9]

  2. Check links and attachments

    Sender domains, URLs, and attachment hashes go to VirusTotal for analysis reports, and Zscaler classifies the URLs and returns Sandbox verdicts for files.[2][10]

  3. Find who acted on it

    The agent checks whether any user reached the URL through Zscaler, then looks for sign-ins after the click in Okta or Entra ID.[2][4][5]

  4. Decide

    Reports that turn out benign or false positive are closed with the evidence attached. Confirmed phishing is escalated for a person to close.[6]

  5. Contain with approval

    If someone entered credentials, the agent proposes revoking their sessions and adding the URL to the ZIA threat blocklist, and asks for approval in Slack.[1][2][7]

  6. Reply to the reporter

    With the Cotool Email tool, the agent tells the employee who reported the email what was found. Each agent can email only the recipients on its own allowlist.[3]

Example

Look up a sender domain in VirusTotalhttp
GET https://www.virustotal.com/api/v3/domains/{domain}
x-apikey: <your API key>

VirusTotal's domain report. Every request carries the API key in the x-apikey header. Indicators queried through the API join VirusTotal's dataset, so don't look up internal or confidential names.[11][12]

What people approve

  • Response agents cannot close an alert as True Positive. When an agent confirms malicious activity, it escalates the alert and a person closes it.[6]
  • Revoking a user's sessions and blocking a URL are state-changing actions that can wait for approval in the app or in Slack.[7]

How Cotool handles it

  • Cotool added Sublime Security ActionMessage analysis for alert context and a Cotool Email tool so agents can reply to employees who reported phishing.[4]

Integrations used

FAQ

Can the agent email everyone who received the phish?
Only addresses on that agent's allowlist. A message can go to up to 50 recipients, and nothing is sent if any recipient is not allowed.[3]

Sources

  1. Respond to every signal (Cotool Respond) · accessed November 16, 2026
  2. Zscaler integration (Cotool docs) · accessed November 16, 2026
  3. Cotool Email (Cotool docs) · accessed November 16, 2026
  4. Changelog (Cotool docs) · accessed November 16, 2026
  5. Microsoft Graph integration (Cotool docs) · accessed November 16, 2026
  6. Alerts (Cotool docs) · accessed November 16, 2026
  7. AI SOC (Cotool) · accessed November 16, 2026
  8. Phishing, T1566 (MITRE ATT&CK) · accessed November 16, 2026
  9. Message groups (Sublime docs) · accessed November 16, 2026
  10. VirusTotal API v3 overview (VirusTotal docs) · accessed November 16, 2026
  11. Get a domain report (VirusTotal API) · accessed November 16, 2026
  12. Authentication (VirusTotal API) · accessed November 16, 2026

Related use cases

All use cases

Related reading

Attackers are scaling with tokens

Cotool helps defenders operate at machine speed. See how security teams are scaling Detection & Response.

Book a demo