A closer comparison would be to Argo Workflow. Tekton and Argo Workflow provide ways to declare workflow pipelines for execution on Kubernetes. Tekton focuses on source based workflows, while Argo is more general purpose.
We've been working with the Argo team to make sure Argo CD works well with Tekton, and we now have a first-class integration in the tekton catalog, contributed by that team.
Disclosure: one of the leads and founders of the Tekton project.
I agree that Tekton and Argo Workflows share a closer comparison than with Argo CD (despite the "CD" acronym appearing in the name Tekton CD). Workflows is aimed to be a more general purpose workflow engine. I like to explain workflows as a fancy Kubernetes Job object, which can be used as a building block to higher level applications.
I believe the majority of workflow users are using workflows for ML, ETL, data processing pipelines. Its strength is being able to leverage your kubernetes cluster as an auto-scaling compute grid, with highly advanced scheduling capabilities (i.e. scheduling based on node attributes, pod priorities, resource requirements, etc...). That said, a fair share of users are using workflows as a lightweight CI solution, though it requires work to glue in other components to provide the complete solution (e.g. Argo Events to handle git webhooks)
Would you normally recommend running tekton in its own k8s cluster, or within the same cluster that runs production workloads? (Particularly if running in something like gke or eks)
We've been working with the Argo team to make sure Argo CD works well with Tekton, and we now have a first-class integration in the tekton catalog, contributed by that team.
Disclosure: one of the leads and founders of the Tekton project.