Dispatch scheduled business logic scans from scan execution policies

Why

With Add business_logic to the scan execution policy... (!269279 - merged), a scan execution policy can declare scan: business_logic on a schedule rule, but nothing runs it yet.
This starts the business logic flow when the schedule fires, as the trigger account that Provision the security policy trigger for sched... (!269920 - merged) provides.

What changes

  • The scheduled run starts one business logic flow per branch, after the policy's CI part.
  • Before each run the account's role is re-synced, and the run is skipped if the account holds more access than the flow.
  • Each skip (flag, license, no flow, provisioning error, missing access) is logged with its reason, and never blocks the CI actions.
  • The trigger's creation is audited before its account's authorizations refresh, as suggested on Provision the security policy trigger for sched... (!269920 - merged).

Design and limitations: https://gitlab.com/gitlab-org/gitlab/-/work_items/638792+

How to test

In a GDK in GitLab.com mode, with Duo Agent Platform onboarding done, the bl_security_analyzer, autonomous_service_account_execution and agentic_analyzer_security_ingestion flags on, and the top-level group opted in to beta features:

  1. Add a scan execution policy with a schedule rule a few minutes ahead and scan: business_logic.
  2. When the cron fires, a security_policy trigger and its service account appear, and a business logic workflow runs as that account.
  3. When it finishes, its findings appear in the vulnerability report.

Verified in a GDK on master 0b951a99, with all four MRs.
The cron fired, and the admin bot created the trigger and a new account.
The scan's CI job succeeded, and its 3 findings matched the project's existing vulnerabilities.

Context

Edited by Meir Benayoun

Merge request reports

Loading
Loading