SkillHub gives teams a private, npm-style registry for AI agent skills, so a useful skill package is published once, reviewed, versioned, and installed into a compatible agent from the command line. Upload a package from the web console, the CLI, or the REST API, pick PUBLIC, NAMESPACE_ONLY, or PRIVATE visibility, and track releases with semantic versions plus tags such as beta and stable. Team namespaces keep their own members with Owner, Admin, and Member roles and publishing policies, while full-text search filters by namespace, downloads, ratings, and recency and hides whatever a user may not see. Every submission enters a review queue: namespace admins approve or reject team skills, platform admins decide promotions to the global scope, and each governance action lands in the audit log. A built-in Skill Scanner inspects published packages and shows its report on the skill page, with optional behavioral, LLM, VirusTotal, and Cisco AI Defense engines. Members can star and rate skills, follow download counts, receive notifications, and report problem packages. Skill suites bundle several skills into one versioned collection with an entry skill, scoped API tokens serve the CLI and scripts, and the interface speaks English, Simplified Chinese, Russian, and Korean. Packages live on local disk or S3-compatible storage such as MinIO. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.