We've noticed that when building skills through Thread SuperMagic, any skill with a large or complex instruction set fails with a network error on the first save attempt. Smaller skills save fine. The only workaround we've found is to build the skill in chunks — save a skeleton first, then use update_skill to add content in passes.

This is friction we shouldn't have to manage manually. The platform should either support larger payloads natively, handle chunking transparently behind the scenes, or at minimum return a clear message telling us why it failed and what to do.

We'd love to see this addressed.....skill-building is a core part of how we're operationalizing SuperMagic, and hitting silent network errors on complex skills slows that down significantly.

**

**