News Summary

• Agent Skills do not “train” agents.
• They soft-fork agent behavior at runtime, telling the model how to perform specific tasks that it may need.
• A soft fork is a backward-compatible change that does not require upgrading every layer of the stack.
• The architecture maps cleanly to how we think about traditional computing.
• You can have dozens of skills available without overwhelming the model’s context window.
• Fine-tuning demands significant investment in talent, compute, data, and ongoing maintenance.