How is the input step intended to be used in a declarative pipeline using a separate block (JENKINS-48379)? There are several issues with this.

• The result of the input can only be used in the current stage. Thus, it is not possible to skip multiple stages or use the input value in multiple stages.
• input and when cannot be combined in all cases. Currently, the result of input can be used in when to skip the stage (cool!). However, if the stage is skipped because of some other criteria, the input is still requested.
• input cannot be used inside the stage steps as a workaround because the agent would be blocked and it cannot be separated to two steps (approval without agent, handling with agent) because the input result is stage-local.

If I missed something and the issues can be solved, it would be nice to know and the documentation should be more clear about this. Currently, it seems impossible to create a declarative pipeline with a manual approval step before a deployment when the process also depends on other criteria (e.g. the branch in a multibranch pipeline).