@dair_ai
If you build agent skills in production, check out this great paper from Alibaba. You can think of a production agent skill as a directory. The root loads on activation, and references, schemas, scripts, assets and nested subskills load only when an execution path reaches them. Compressing the root alone misses most of the deployment cost, and it can push branch-specific detail into context that is always loaded. SkillZip Pro compresses the whole bundle, removing content from a reference or subskill when the root or a declared environment contract already provides it, while preserving routing so every required file and directly callable entry stays reachable after the rewrite. On a production content-moderation skill it removes 38% of bundle tokens and 10.4% of end-to-end per-run tokens with no quality loss. Four modes cover the deployment cases: - One-Shot rebuilds the bundle - Continual applies Zip-on-Write after each evolution patch - Persistent rewrites the shipped bundle to cut storage and runtime context, and - Transient keeps the shipped bundle byte-identical while building a task-specific view Paper: https://t.co/ZpPCQh0I9J Chat with Paper: https://t.co/zgjSp5CXgW