Currently, it is risky to create new scripted fields: a mistake on the code can cause discovery/visualization and dashboards to break, even when the new field is not used.
Also there is not much opportunities to test the code in Kibana before "deploying"/"saving" it.

I think Kibana needs a set of features to reduce the risk of breaking, but also empower users while coding/testing new scripted fields.

I can think of:

• run the code against some rows and see the output at script writing time: so there would be a test section in the form where you can provide a search to narrow the rows you want to test your script against.

• script field sandbox/isolation: if a field is broken or failing, fail just what uses the field. For instance, if in discovery the field is selected, everything is shown but that field column with some error icon. Visualizations that use the field for aggregating or filtering show a error in the widget, but a complete failure on the whole dashboard on every dashboard.

• syntax highlighting and API auto completion.