These days I rarely write pipelines. I mostly write specs, an agent dumps 500 lines of YAML, Python and a Makefile then I get my artifacts. The amount is overwhelming and a liability, and I stop trusting what I just merged. So I pivoted my CI/CD project toward more visualization, analytics and data aggregation, so engineers can see what agents and the pipeline produced, the flow, steps, data. Where it fails. With that I have a better mental model instead of a blob of code.
The goal is better observability, and pipelines that explain their own failures back to the next agent. And also with a personal focus on UX and design. My project supports Perforce P4, Git/GitHub and GitLab, so quite flexible.
Would love to hear your impression.
I'm also very interested in your CI/CD horror stories, and how you use them these days. Any input is much appreciated!
P.S. Code is public and provides GitHub Attestations.