Sublation V5.0
A local-first governance layer for AI-agent skills that adds a multi-agent collaboration panel, a routing system that only suggests which skill fits a request, and a schema for controlling how skills are promoted, without granting permissions or executing anything itself.
Sublation V5.0 is a Simplified Chinese release package that adds skill governance on top of a multi-agent setup: a collaboration panel for several agents, a skill-classification and shadow-routing system that suggests which skill likely applies to a request, and a governance schema for managing that process, together with synthetic test fixtures and a regression test suite.
The package is explicit about its boundaries: a panel seat is not review authorization, a queued or woken intent does not mean an agent has replied, and the shadow router only suggests a skill rather than executing it, changing the host's own choice, or granting any permission. Automatic wake behavior is disabled by default, and the package does not install, register or enable any watcher, scheduler or real host adapter on its own; writing to a formal skill root, promoting a candidate, sending external messages or uploading material all require explicit user approval first. Installation is unzipping a single top-level directory into the target agent's skills folder, and the full regression suite is run with Python's unittest module.