Security feedback is easiest to use while the person making a change still understands its context. A finding that arrives after merge or near a release can require someone to reopen work, reconstruct a decision, and coordinate another delivery. Early feedback shortens that loop.
Central oversight makes that feedback dependable across projects. Security owners define the risks that require action, platform owners maintain shared controls, and application teams resolve problems in their code. Developers need a usable route through that process, including a way to question a false positive or an unsuitable rule.
Bring findings into the work being reviewed
Run appropriate scans in merge request pipelines and give the developer enough information to act: the affected file or package, the nature and severity of the finding, and guidance on how to fix it. GitLab's integrated security reports can show this in the merge request.1 Teams working directly from scanner artifacts need an equally visible way to bring results into review.
Show which problems the change introduces and which were already present. Check that the source and comparison pipelines produce the reports the review depends on. A missing or failed scan needs attention as a coverage failure; it can't be interpreted as a clean result.
If your scans run but their results aren't reaching review, establishing security scan coverage explains how to follow a finding from the job to the developer's next decision.
Let developers run checks and use remediation playbooks without waiting for another team. Keep guidance specific to the languages and frameworks they use. For example, explaining why parameterized queries avoid SQL injection gives developers a way to address the problem. A generic instruction to improve security leaves them guessing.
Make the security contact and exception path easy to find. A developer who understands the finding but can't get a risk decision is still blocked. Agree on response times that reflect the urgency of the work, and make clear who can decide when the usual reviewer is unavailable.
Roll out policy with the teams affected
Try new scanner settings with representative applications before applying them broadly. Confirm that checks run in the intended branch and merge request pipelines, findings are understandable, and the runner fleet can support the additional work. Large repositories and restricted networks can reveal problems that a small demonstration project won't.
Shared configuration makes adoption easier, but projects can change or remove ordinary includes. Once a requirement is agreed, a security policy project can apply centrally managed policies to the intended groups and projects.2 Check both policy scope and actual pipeline behavior as coverage expands.
For a new merge request approval policy, warn mode lets teams examine its effect before making it fully blocking. The policy shows violations, and a user can bypass them by providing a reason. Other merge requirements still apply. Review the warning volume, bypass reasons, and time spent resolving findings with the affected teams. Use those observations to judge whether the rule catches the intended problems or needs tuning.
Keep existing mandatory protections in effect during that trial. Warn mode is a rollout choice for the policy being evaluated, not an indication that its violations have been resolved. Keep the bypass reasons with the finding so the person reviewing risk can consider them.
Security policy from observation to enforcement develops that approach, from choosing the right policy type to expanding it without losing the ability to correct a rollout problem.
Improve both the rule and the response
Measure how long it takes to get from a finding to an informed decision and then to a fix. Scanner duration is only part of that time. Unclear messages, unavailable reviewers, and uncertainty over ownership can create longer delays than the scan itself.
Ask developers about their actual experience and compare it with the view of security and platform owners. A centrally maintained rule can be technically correct yet difficult to follow because the instructions don't fit a team's workflow. Repeated questions and investigated false positives can guide changes to scanner settings, examples, and training.
Central ownership works best when teams can propose improvements and see what happens to them. Keep the security requirement clear, explain exceptions, and confirm that a policy change has reached the intended projects before relying on it.
Further Reading
Notes
-
Integrated merge request security reports and the vulnerability report are available in Ultimate. Without those views, teams need another way to bring scanner results into review and track findings to resolution. ↩
-
Security policies, including merge request approval policies and their warn mode, are available in Ultimate. They let policy owners manage required checks across projects. ↩
- Understand security results in pipelines and merge requests explains the reports developers and reviewers use.
- Set up security policy projects covers where policies live and how they apply to projects.
- Pilot approval policies with warn mode explains violations, bypasses, and the transition to enforcement.
- Check approval policy pipeline requirements helps investigate missing results or an unexpected approval requirement.