Agents & Skills
This section covers reusable agent behavior: how skills are designed, how ready-made skills are organized, and how policies constrain agent behavior across repositories and tools.
Skills versus tools
A tool exposes a capability such as reading a repository, querying documentation, or running a validation command. A skill defines a reusable workflow for deciding when and how to use capabilities to achieve a specific class of task.
What belongs here
- skill design and authoring conventions;
- ready-made reusable skills;
- agent policies and behavioral constraints;
- guidance for testing and versioning reusable agent behavior.
Reusable skill assets belong under agent-assets/skills/. Policy assets belong under agent-assets/policies/. Tool-specific setup remains under AI Tools or MCP.