mxs skill is the canonical entry point for downstream AI agents. The chapter set covers commands, authoring workflows, liteXML, target selection, output modes, and safety. Use mxs skill to enumerate, mxs skill get <slug> to load one chapter, mxs skill all for everything, and mxs skill search <kw> for substring lookup. Default output is raw markdown (--output llm) suitable for context injection.
When changing CLI surface, update the corresponding chapter under packages/cli/skills/*.md in the same change set so agents see current behaviour.
When implementing or completing user-facing mxs CLI functionality, update packages/cli/README.md in the same change set before considering the task complete.
- Add newly implemented commands, flags, output modes, authentication behavior, configuration behavior, and file formats to the README.
- Prefer concise updates under existing command, option, or capability sections instead of creating broad new sections.
- Do not add documentation churn for purely internal refactors, test-only changes, mechanical dependency updates, or bug fixes whose observable CLI behavior remains unchanged.
- When a feature affects both human and AI-agent usage, document the machine-readable and readable output contracts explicitly.