腾讯SkillHub

What Tencent SkillHub is

Tencent SkillHub is a skills community and store optimised for Chinese-speaking users, connecting skills to an agent client so that capabilities can be reused, and in some cases paid for per call. It is the most institutionally built-out entry in this group, and it is interesting precisely because it treats skills as a marketplace with real admission requirements rather than a directory of links.

What the platform provides

You browse a plugin plaza to discover and install open-source plugins, explore expert packs that chain several skills into an end-to-end deliverable, view certified companies and official skills in an enterprise zone, and publish your own through a web console, a command line, or agent-driven deployment with version history and rollback. The loop runs discover, install, publish, monetise.

A per-call billing service lets enterprises attach paid services to agent tasks and settle by invocation, which is a different commercial model from the subscriptions common elsewhere. Trending and download rankings surface what is gaining traction, and the expert packs are the most useful packaging here, because chaining several skills into one deliverable is what most users actually want.

Who it is for

Chinese-speaking developers, and enterprises that want to publish, govern or monetise agent capabilities. The enterprise zone and the publishing console are clearly aimed at organisations rather than individuals.

What to keep in mind

Start with the numbers, because the platform says so itself: the headline usage and earnings figures on the page are labelled as sample data rather than real metrics, which is honest and means they should not be read as traction. The billing model is per call for enterprise services, but no rate, revenue share or settlement period is published, so the economics have to be established directly before anyone builds revenue assumptions on it.

The admission pipeline is the strongest feature and the one worth weighing properly. Entry to the store is described as mandatory and thorough: content compliance filtering, a deep vulnerability scan by the vendor's own security lab, and an AI model safety assessment by a second lab, with real-name verification through the vendor's face-check service and encrypted storage of that identity data. Layered review by named labs is more than most of this category offers, and the identity verification is a genuine gate for publishers. It also means publishers hand over biometric identity data, which is a category of personal data with its own obligations and breach consequences — worth understanding before you register.

One operational point needs flagging. The site instructs users to copy an external markdown link and hand it to their agent to perform the installation. That asks an agent to act on the contents of a file nobody in your organisation has read, which is exactly the pattern prompt injection relies on. Download and review the file yourself before letting an agent run it.