How I added a plain-words phishing warning above suspicious mail, why it only warns and never blocks, and what customising your mail client can do for safety and speed.
Most of the risky mail I receive does not look risky. It looks like a renewal notice, a shared document or a payment that failed. A few weeks ago I added a warning of my own to the shared inbox I use for my projects: a red bar across the top of a message that says, in plain words, why the mail might not be what it claims to be. This post explains what the bar checks, why it only warns and never blocks, and why I think customising the tool you read mail in is one of the cheapest safety and efficiency wins available. I run it every day, so the lessons below come from real mail, not from a test lab.
Why the inbox is the weak spot
Spam filters are good at volume. They are weaker at the single, careful mail that is written for you. Those mails often pass the filter because, technically, nothing is wrong with them: a real free-mail account sent them, the links work and the wording is polite.
The information that gives them away is usually already there. It is just hidden. Your mail server records whether the sender passed SPF, DKIM and DMARC, the three public standards that let a domain say which servers may send mail in its name. Most mail clients bury those results in the raw headers, where nobody looks while answering twenty mails before lunch.
So the job was not to invent new detection. It was to take signals that already exist and put them where my eyes are.
What the warning bar does
When a new mail arrives, a small extension in my inbox scores it once and stores the result with the message. If the score is high enough, a red bar appears directly under the subject. Each reason is one short sentence, for example:
- Sender name versus sender address: the name says a well-known company, but the address belongs to a free-mail service.
- Authentication results: the sending domain failed its own DMARC check, or a mail claims to come from my own domain without passing SPF.
- Links that do not match: the link text shows one website, but the link goes somewhere else.
- Links to anyone-can-host storage: a "view invoice" button that points to a file anyone could have uploaded.
- Known bad links: a link host appears on a public list of reported phishing sites, refreshed daily.
- Pressure wording: phrases such as "payment failed" or "verify your account", which on their own mean little but add weight next to the rest.
No single signal decides. A newsletter can have mismatched links, and an honest freelancer can use a free-mail address. The bar appears only when several signals add up, and it always shows all of them, so I can judge for myself in two seconds.
Warn, do not block
The most important design choice was what the extension is not allowed to do. It never moves, tags, closes, deletes or blocks a message. It only adds the bar.
That choice comes from the cost of a mistake. A missed warning means I read a suspicious mail with my normal care. A mail that is wrongly blocked can be a real customer who never gets an answer, and I would never find out. A warning that is sometimes wrong is cheap. A filter that is sometimes wrong is expensive and silent.
It also keeps me in the loop. Because every reason is written out, I learn the patterns over time. After a week the bar was teaching me as much as it was protecting me.
The trap with "no result"
One lesson deserves its own note. Some legitimate mail, such as the messages my own website's contact form sends, arrives with no authentication result at all, because it never leaves my own systems. My first version treated "no result" like "failed" and lit up a red bar on every contact form message. The rule now only fires on an explicit failure. Missing evidence is not evidence of fraud.
How I tested it before switching it on
Before the bar went live I ran the scoring over a large batch of older, already-handled mail, in read-only mode, and looked at every message that would have been flagged. The goal was simple: no real customer or partner mail with a red bar. Where a legitimate sender tripped a rule, I fixed the rule or added that sender to the right allowlist, then ran the test again. Only when a full run came back clean did I turn it on for new mail.
I repeat that test after every change to the rules. It takes minutes and it is the only reason I trust the bar.
Customising the client for efficiency too
The same idea, put the useful thing where you already look, works just as well for speed as for safety. Once you accept that your inbox is software you can extend, small additions add up:
- Blocking tracking images: remote images in incoming mail stay off until I choose to load them, so senders cannot see when or where I opened their message.
- Customer notes above the thread: a short note about the sender shows at the top, so I do not have to search for context before replying.
- Related conversations: mail from the same sender in other inboxes is listed next to the message.
- Reply formatting and status rules: saved replies keep a consistent layout, and a conversation stays in the state I left it in.
None of these is big on its own. Together they remove dozens of small clicks and a few real risks every day. FreeScout, the open-source help desk I use, accepts modules, and that makes this realistic for one person: each addition is a small, separate piece I can switch off in one click if it misbehaves.
What I do not know yet
The rules are tuned on my own mail, which is mostly English and about my own projects. I do not yet know how well they catch lures in other languages, such as fake parcel or bank messages in Dutch. I also do not know how attackers will adapt once more inboxes show this kind of warning. A phishing mail sent from a fully authenticated, freshly registered domain with clean links will still pass, and only my own attention catches that one.
Frequently asked questions
Does a warning bar replace a spam filter?
No. The spam filter removes bulk junk before I see it. The bar is a second layer for the careful, targeted mails that get through, and it relies on the filter's own authentication results.
Why not move suspicious mail to a separate folder automatically?
Because a wrong decision would be invisible. A legitimate mail moved away might never be answered. A bar on a real mail costs me a second; a hidden real mail can cost a customer.
Do I need to write code to customise my mail client like this?
For a scoring extension, yes, a little. Many improvements need much less: switching off remote images, adding sender notes and using saved replies are often settings or ready-made add-ons. Start with those and write code only for the gap that remains.
What to do first
- Open the raw headers of a few recent mails and find the SPF, DKIM and DMARC results, so you know what your server already tells you.
- Check that your own domain publishes SPF and DMARC records and that every service sending as you is included.
- Turn off automatic loading of remote images in your mail client.
- Pick the one check you would most like to see at a glance, and find out whether your client can show it above the message.
- Test any new rule on old mail before it touches new mail, and keep it warn-only.
If you are curious about the projects this inbox serves, Blockspot.io (opens in a new tab) is the main one, and you can follow smaller changes like this one on my updates page.