This is indeed a bit of an anti-pattern, although other people have done similarly. The best way to do it would be using an

after_context_created

hook like this: https://github.com/kedro-org/kedro/discussions/1436#discussioncomment-2789761 If it suits your use case, better would be to register two pipelines in `pipeline_registry.py`:

pipeline_1 = pipeline(...)
pipeline_2 = pipeline_1 + [node_a]

and then run the pipeline you want through

kedro run -p

Another alternative is to keep the node in the pipeline but just change the logic inside the node function itself to skip using something like on

if not do_step_a: return