diff --git a/keps/draft-20190418-skeleton-generator.md b/keps/draft-20190418-skeleton-generator.md
new file mode 100644
index 000000000..b9ed3c7e6
--- /dev/null
+++ b/keps/draft-20190418-skeleton-generator.md
@@ -0,0 +1,39 @@
+---
+kep-number: draft-20190418
+title: Skeleton generator
+authors:
+  - "@djannot"
+owners:
+  - TBD
+editor: TBD
+creation-date: 2019-04-18
+status: provisional
+---
+
+# Skeleton generator
+
+## Table of Contents
+
+  * [Table of Contents](#table-of-contents)
+  * [Summary](#summary)
+  * [Motivation](#motivation)
+
+Created by [gh-md-toc](https://github.com/ekalinin/github-markdown-toc)
+
+## Summary
+
+In order to develop a new framework using KUDO, you always need to create several `Plans` (`deploy`, `update`, `upgrade`, ...) and for each plan to create several `Phases`, `Steps` and `Tasks`.
+
+Also, any Framework will have a `Task` in the `deploy` plan to create similar Kubernetes objects (`StatefulSets`, `Services`, ...).
+
+A Skeleton generator could create `yaml` templates with the common `Plans`, `Phases`, `Steps` and `Tasks`.
+
+This generator should be part of the KUDO CLI.
+
+With the new proposed approach based on dynamic CRDs, these elements will be defined in separated `yaml` files, so it will be even easier to create a Skeleton generator with a clear directory structure.
+
+## Motivation
+
+Having a Skeleton generator would help people getting started faster when they want to develop a new KUDO framework.
+
+It would also make the framework created using this generator more readable as they would follow a naming convention for the different objects (`StatefulSets`, `Services`, ...).