UZE, the official UZE plugin
This is a normal Agent Plugins 1.0 package. It carries no special treatment in the Store, the router, or any integration: it is installed, discovered and delivered exactly the way any other Skill-only package is.
It ships six Skills:
skills/init/SKILL.md: an agentic orchestrator that calls UZE's own deterministicuze agent context inspect|plan|reconcileCLI to make a project's instructions context portable.skills/worktree/SKILL.md: coordinates isolated worktrees for concurrent agent work and safe integration, honouring theworkspace:policy the project declares inagents.yaml.skills/rules/SKILL.md: writing a project's rules under.agents/rules/, the instructions that arrive only where they are relevant, proven withuze agent rules check.skills/architect/SKILL.md: writing the Mermaid diagrams a project keeps underworkspace.artifacts: which view a change belongs in, how the architect surface lists it, anduze agent artifacts checkas the proof it draws.skills/author/SKILL.md: creating a plugin end to end, from choosing or creating its marketplace to checking, installing and iterating on it;references/extension.mdbeside it covers writing a declared extension the plugin carries: its manifest, a script or a program, its grants, its tests, and the consent and enabling it needs before it runs.skills/try/SKILL.md: making a project to try the marketplace's example plugins in, withdemo.pybeside it writing the project (a Git history and a GSD milestone in.planning/), and installing the examples there.
The plugin format (skills, agents, MCP servers, hooks, and the
${PLUGIN_ROOT} placeholder) and what each harness receives are documented
at uze.sh/docs/reference/plugin-format
and uze.sh/docs/reference/harnesses.
See docs/capabilities/context-manager.md
for the architecture these sit on top of, and
docs/capabilities/uze-skill.md for how
they are invoked per harness, the no-special-treatment proof, and what is not
tested.