On June 8th, Xiaohongshu officially launched its RED Skill feature.

When some creators publish notes, they can attach an AI Skill component below the content. Users can click to copy a command and install it for use within Agent products that support the relevant capabilities. The platform has also launched official support activities and a curated Skill ranking list.

Viewed from a broader timeline, this is not an isolated product update.

Over a month prior, on April 30th, Xiaohongshu issued a company-wide internal letter announcing the establishment of Dots, a first-tier department with AI at its core.

This department does not report to any specific business line but reports directly to the new president, Conan. It possesses independent budgeting and personnel authority, standing alongside core businesses such as community, e-commerce, and commercialization. Concurrently, Xiaohongshu also established an Enterprise Intelligence Department, preparing for the AI era across organizational, technological, and efficiency dimensions.

Over the past few years, Xiaohongshu's moves in AI have been relatively restrained, focusing more on scenarios like content production, search, and recommendation. However, with the establishment of Dots, AI has begun to elevate from a tool capability to a company-level strategic direction.

RED Skill represents a concrete implementation of this shift on the community side.

The problem it addresses is straightforward: lowering the barrier to discovering and disseminating AI Skills.

For the better part of the past year, Skills have gradually become one of the most active topics in the Agent field.

Simply put, a Skill is a structured set of operational instructions for an AI Agent to invoke. When an Agent reads these instructions, it can complete specific tasks according to predefined workflows.

For example, a seasoned product manager could encapsulate their method for writing requirement documents into a Skill. Once installed by other users' Agents, they could also complete the work following a similar process.

As the number of Skills continues to grow, how to enable users to discover them has become a new challenge.

Initially, this role was primarily filled by developer communities like GitHub and ClawHub. Entering 2026, an increasing number of Agent platforms have begun building their own Skill ecosystems and capability markets. Major players including Tencent, Alibaba, and ByteDance have all launched relevant capability entry points for developers and users, with companies like Zhipu AI and Meituan also following suit.

Compared to these platforms, Xiaohongshu's entry point is not entirely the same.

Most Agent platforms attempt to build Skill execution ecosystems, whereas Xiaohongshu's approach is closer to a content community's perspective on Skill discovery and dissemination scenarios.

Understanding this point may make it easier to comprehend the rationale behind RED Skill's emergence.

As Agents gradually become a new gateway for information processing, the discovery, dissemination, and installation pathways surrounding Skills are beginning to form new traffic scenarios. Xiaohongshu possesses a relatively unique resource in its hands—a continuously growing technical content community.

This was not formed overnight.

According to officially disclosed data, over the past year, the volume of technology content published on Xiaohongshu has increased by over 100% year-on-year, and the scale of creators has grown by more than 200% year-on-year.

Content related to "Build in Public" on the platform has accumulated to over 1.1 million entries, with those born in the 2000s and 2005s becoming the main participants.

Simultaneously, the number of active developers on the platform has exceeded 160,000, with over 90% of them having developed more than one product within a year.

Regarding the Skill direction, Xiaohongshu disclosed that the number of creators publishing AI Skill-related content on the platform has reached 300,000, with related topic exposures exceeding 600 million times. Since the launch of RED Skill, nearly a thousand original Skills have been published within the community.

Beyond these numbers, the change in demographic structure is more noteworthy.

In April of this year, Xiaohongshu hosted its first Hackathon Summit, where 200 shortlisted developers engaged in 48 hours of closed-door development. Over 60% of them were born in the 2000s, with the youngest participant being only 13 years old.

Compared to the previous generation of developers, these AI natives are more accustomed to utilizing AI for development work and are more willing to publicly share their product-building processes.

Project progress, product inspiration, user feedback, and even failure experiences have all become part of the public discussion.

This culture, known as "Build in Public," is becoming a significant characteristic of Xiaohongshu's technical community.

Sanbing, the head of Xiaohongshu's technology vertical, once summarized it as: "enabling creators to showcase progress, entrepreneurs to find partners, and investors to discover projects."

From this perspective, RED Skill is more akin to adding a standardized dissemination pathway for Skills on top of the existing content ecosystem.

Of course, the current RED Skill still has obvious limitations.

After users discover a Skill on Xiaohongshu, the subsequent installation and execution processes still require redirection to external Agent products to complete. Xiaohongshu cannot directly perceive whether users have truly completed the installation, whether they continue to use it, or what the actual performance is like.

Another detail is that the "number of users" displayed on the detail page currently counts those who clicked the "Use" button, not the number who ultimately completed the installation or actually invoked the Skill.

This means there is still significant conversion loss between discovery and usage.

At this stage, RED Skill is closer to a discovery gateway for Skills rather than a comprehensive Skill ecosystem platform.

However, as Skill formats gradually become more standardized and compatibility between different Agents improves, the methods of disseminating Skills are still likely to evolve.

It remains unclear exactly what role content communities will play—whether as traffic gateways, product marketplaces, or developer communities.

One thing is certain: as the Agent ecosystem continues to develop, the discovery and dissemination of Skills are becoming a new competitive frontier.

RED Skill is just the beginning of this change. Who will truly capture users in the future will still depend on whether a complete closed loop can be formed across the stages of discovery, installation, execution, and retention.

And Xiaohongshu is attempting to secure a position within part of that loop.