A delivery process is easier to trust when teams can explain how a change reaches production, what access each job receives, and how they'll restore service if the change goes wrong. Fast pipelines help, but predictable changes and a recovery path matter just as much.
Shared pipeline code makes improvements available to many projects. It also spreads mistakes, so ownership, representative tests, versioned releases, and a manageable upgrade path belong together. Standardize the work projects have in common while leaving room for differences in build tools, release cadence, and operating needs.
The code a pipeline runs determines the trust it needs. Unreviewed contributions, protected builds, signing, and production deployment shouldn't gain the same access simply because they use the same CI/CD platform. Runner selection, host isolation, network access, and job credentials each control a different part of that risk.
A delivery record should connect the deployed artifact to its source, pipeline, environment, and result. Recovery adds another question: can the team restore the service, including the data and configuration it depends on? Testing that path before an incident makes release decisions more informed.
The Design principles develop these choices and their tradeoffs. Recommendations offer practical approaches to the pillar's outcomes. Use the CI/CD and release checklist to find where your own delivery process needs attention, starting with an ordinary change or an urgent fix.