Phishing-report inbox triage for small teams
Sort emails staff forward as suspicious, summarize the warning signs, and queue them by urgency for the person who handles security.
The use case
Small teams often set up a shared address where staff forward suspicious emails, but it fills up with a mix of real phishing, spam, and legitimate messages nobody recognized. The one person responsible for security may not check it for days. This agent reads each forwarded report, extracts sender details, links, and attachment names without opening them, notes common warning signs, and sorts reports into likely phishing, likely spam, likely legitimate, or unclear. It drafts a short thank-you note to the reporter and an urgency-ranked summary for the security contact. That person confirms the verdict and decides on any blocking, cleanup, or staff alert.
Example workflow
- A staff member forwards a suspicious message to the shared report address, which triggers the workflow.
- The agent extracts sender address, reply-to, display name, link domains, and attachment names as text without clicking or opening anything.
- It compares these against warning signs you list, such as lookalike domains or urgent payment requests, and notes which apply.
- Each report is labeled likely phishing, likely spam, likely legitimate, or unclear, with reports mentioning clicks or entered credentials marked urgent.
- The security contact reviews the queue, confirms or changes labels, and decides on blocking, password resets, or a team warning.
- The agent drafts a reply to the reporter for the security contact to send, and each case is logged with its final outcome.
Tools
- Google Workspace or Microsoft 365 shared mailbox
- Claude
- Slack
- Google Sheets
- Make
- Email security console
Risks and guardrails
- Opening links or attachments could trigger malware — guardrail: the agent handles links and filenames as plain text only and never fetches or opens them.
- A real phishing email could be labeled legitimate — guardrail: likely legitimate is still reviewed by the security contact before the case is closed.
- A staff member who clicked a link may be waiting too long for help — guardrail: any mention of clicking or entering credentials is marked urgent and sent to the security contact immediately.
- Blocking a sender or deleting mail across inboxes is disruptive — guardrail: the agent takes no blocking or deletion actions; the security contact decides and acts.
Related recipe
Support Ticket and Review Triage with Policy-Cited Draft Replies: An agent tags every helpdesk ticket and public review by intent, urgency, and sentiment, routes it, and drafts a reply that cites your help center. People approve refunds and all public replies.
Want it set up for you? Request a quote.
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Google Workspace extension (official page): github.com/gemini-cli-extensions/workspace (checked 2026-11-19)
- Claude (official page): claude.ai (checked 2026-11-09)
- Slack MCP server (official page): docs.slack.dev/ai/slack-mcp-server (checked 2026-11-09)
- Make MCP server (official page): developers.make.com/mcp-server (checked 2026-11-09)