The problem: Security adds a code scanner to the build. On day one it fails builds because of hundreds of old problems and false alarms. Developers say, “The scanner stops us from shipping.” Security says, “The scanner is doing its job.” Soon the developers skip it or turn it off.
Words used here: a build is the step that turns the developers’ code into a working product. A scanner is a tool that reads code and looks for known weaknesses. A false alarm is a warning about something that is not really a problem.
Think of a smoke alarm
A smoke alarm in the kitchen rings every time you make toast. The first week you wave a towel at it. The second week you take the battery out. Months later there is a real fire in the pan, and the alarm is silent. The alarm was not wrong to exist. It was set up so that people could not live with it.
A noisy scanner gets the same treatment.
What it looks like at work
Security adds the scanner to the build, because a client asked for code scanning. The first run reports hundreds of findings, most of them in old code, and many of them not real. Every build now fails. A release is stuck two days before a client demo. The lead developer adds a setting that lets the build continue anyway. Nobody tells security. From then on, the scanner runs, and nobody reads it.
- 1Scanner goes onClient asked for code scanningDay 1: all blocked
- 2First build412 findings. Build failed!Mostly old code
- 3Demo in 2 daysRelease is stuck, demo is FridayRelease stuck
- 4Quick switch-offI added a setting to skip the scanSecurity not told
- 5Months laterScanner runs, but nobody reads itNew flaws ship unseen
Both sides are half right. Security is right that code must be scanned, because a client or auditor will ask. The developers are right that a scanner that stops every release cannot stay on. What is missing is a gentle start and an agreed rule for what may fail a build.
You will notice it when:
- Builds failed on the first day the scanner was turned on.
- Developers say “just ignore the scanner.”
- The scanner has a setting that lets the build pass anyway.
- Nobody owns the list of old findings.
- The scanner has never been tuned for this team’s code.
What exactly needs fixing
The first fix is not a better scanner. It is a start that does not block people, and a clear rule on what is allowed to stop a build. This card shows the whole problem on one page:
Code scanner stops builds, so it was switched off
Build pipeline of the Orders app. A skip setting, security not told.
A serious new flaw is released while the scanner is muted.
- Release delays each time it is on
- Bad audit result, worse than none
- Developers stop trusting tools
- Real flaws ship unseen
- 1Start report-onlyThe scanner shows findings but never fails a build
- 2Block only the newNew critical findings fail. Old ones: dated clean-up
- 3Tune, then tightenRemove false alarms with a developer, add rules later
Dev lead owns the build rules. Security owns the scanner.
Report-only for the first weeks, blocking date agreed in writing.
A blank copy of this card is in the Tools section.
Why it happens
- The scanner starts in “block everything” mode. Nobody planned a gentle start.
- Old problems and new problems are mixed together. A developer is blocked by code they did not write.
- False alarms are not cleaned. After a few, people stop trusting the tool.
- The developers were not asked. The scanner appeared in their build overnight.
- A client asked for scanning. So the goal became “scanner is on,” not “code is safer.”
- Nobody checks whether the scanner is still on. A silent bypass can last for months.
What happens if you ignore it
You have a scanner on paper, and no real protection.
- Real new weaknesses are released, because nobody reads the report.
- The client or auditor asks, finds the bypass, and the finding is worse than having no scanner.
- Developers learn that security tools can be ignored, and the next tool gets the same treatment.
- Release dates slip each time the scanner is switched back on.
At a glance
| Question | Answer |
|---|---|
| How likely is it to hurt us? | Very likely, when a scanner is added without a plan. |
| Worst case? | A serious flaw is released while the scanner was muted. |
| What does it cost? | Delays, a bad audit result, and lost trust between the teams. |
| What if we wait? | The bypass becomes normal, and the old findings pile up. |
| What does the fix cost? | A few days of tuning and a short agreement with the developers. |
How to fix it, step by step
Bring the scanner in gently, and make it trusted before it is strict.
- Start in report-only mode. The scanner runs and shows findings, but never fails a build, for the first few weeks.
- Separate old findings from new ones. Put the old findings on a dated clean-up list. Only new problems matter for the build.
- Remove the obvious false alarms. Sit with a developer and tune the rules. A tool with fewer, better warnings gets read.
- Agree what may fail a build. At first: only new critical findings. Write it down with the dev lead.
- Tighten step by step. Add more rules as the noise goes down and the team gets used to the reports.
- Make a bypass visible. If a build passes with the scanner skipped, security is told the same day.
- Count the noise. Track how many warnings were real. Fewer false alarms means more trust.
The picture shows the three stages. Never jump straight to the last one.
Who to bring in: The dev lead (the build rules), security (the scanner and the findings), the team that owns the pipeline, and the client contact if scanning is in the contract.
Tools and templates
What may fail a build, and when
| Stage | The scanner may fail the build for | Old findings |
|---|---|---|
| 1. Watch | Nothing. Report only. | Listed, with a clean-up plan |
| 2. Block the new | New critical findings | Cleaned by a set date |
| 3. Tighten | New critical and high findings | Cleaned by priority |
Blank finding card
| Field | Write here |
|---|---|
| What is broken | One line, in plain words |
| Where it is | The system, team, or process |
| What could go wrong | In one sentence |
| What it costs us | Money, time, a leak, lost trust |
| The fix | The exact action |
| Who fixes it | One name or one team |
| By when | A date, not “soon” |
Email to the dev lead
Subject: A gentler start for the code scanner
Hi [name]. I know the scanner stopped your builds last week, and I understand why it was a problem. I would like to put it in report-only mode for a few weeks, so it never blocks a release. We will go through the false alarms together, and only turn blocking on for new critical findings. Can we meet for 30 minutes this week to agree the rules?
Field tip
A scanner people trust and read is worth more than a strict one they have switched off.
Words used in this chapter
- Build: the step that turns the developers’ code into a working product.
- Scanner: a tool that reads code and looks for known weaknesses.
- False alarm: a warning about something that is not really a problem.
- Pipeline: the chain of automatic steps that builds, tests, and releases code.
- Bypass: a setting that lets a build pass without the check.