Anthropic announces donation of Agent Skills standard to the Linux Foundation AAIF
BIBIBI
AT A GLANCE
Anthropic announces donation of Agent Skills standard to the Linux Foundation AAIF, with joint maintenance by engineers from multiple parties.
Article
Anthropic is also donating Skills to the Linux Foundation, following MCP its earlier handover of another set of Agent standards
Beating AI News flash: Anthropic announced that it will donate the Agent Skills standard to the Linux Foundation's Agentic AI Foundation（AAIF). Anthropic designed this format while developing Claude Code, 2025 October launched it,December and opened it as a cross-platform standard. ChatGPT, Cursor, GitHub Copilot, and more than 70 other Agent platforms have already integrated it.
A Skill is equivalent to preparing an Agent with a reusable "work manual." Developers write task steps into AI Agent SKILL.md, and can also attach scripts and reference materials, allowing different Agents to read them as needed.MCP is responsible for connecting Agents to external tools and data, while Skills teach them how to complete specific tasks.
The donation covers the standard's name, branding, GitHub organization, and official website domain.AAIF was jointly initiated in 2025 by Anthropic, OpenAI, and Block, and is housed under the Linux Foundation; it had previously received MCP. Going forward, engineers from Anthropic, OpenAI, and Google will jointly maintain the Skills standard in their personal capacities, rather than Anthropic continuing to lead it alone.
The Skills ecosystem is expanding rapidly. Vercel reports that the skills.sh directory lists approximately 1000000 skills, with nearly 280000000 installations in total. Security issues are also beginning to emerge: Snyk previously confirmed 76 malicious skills among samples from ClawHub. Anthropic has proposed establishing version management, a formal proposal process, and security guidelines.
Original article link https://m.theblockbeats.info/flash/373319
Key points
01
The report states that Anthropic will donate the Agent Skills standard to the Linux Foundation's Agentic AI Foundation（AAIF）。
02
The format was designed by Anthropic while developing Claude Code, and was launched in 2025 October and December opened as a cross-platform standard.
03
The report claims that ChatGPT, Cursor, GitHub Copilot, and other 70 Agent platforms have already integrated it.
04
A Skill uses SKILL.md, along with optionally attached scripts and reference materials, to provide AI Agents with reusable task steps.
05
MCP is responsible for connecting external tools and data, while Skills guide Agents in completing specific tasks.
06
The donation includes the standard's name, branding, GitHub organization, and official website domain.
07
The report states that engineers from Anthropic, OpenAI, and Google will jointly maintain the Skills standard in their personal capacities going forward.
08
Vercel reports that the skills.sh directory lists approximately 1000000 skills, with nearly 280000000 installations in total.
09
Snyk previously confirmed 76 malicious skills among samples from ClawHub.
10
Anthropic has proposed establishing version management, a formal proposal process, and security guidelines.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
Skills are like a work manual prepared for AI Agents: developers write methods for completing tasks into SKILL.md and may attach scripts and materials. This arrangement means that the relevant standard will be transferred to AAIF under the Linux Foundation and jointly maintained by engineers from multiple parties, rather than continuing to be led solely by Anthropic.
Why it matters to readers
Reports indicate that Skills has been integrated by multiple Agent platforms and has a large skills directory and installation volume; meanwhile, malicious skills have already appeared, so subsequent governance, version management, and security guidelines warrant attention.
Beginners can use it to understand two types of components in the Agent ecosystem:MCP helps Agents connect to tools and data, while Skills tell Agents how to perform specific tasks; security risks should also be noted when using third-party skills.
Risks and unknowns
The evidence consists only of a media news flash and does not include a direct announcement from Anthropic, the Linux Foundation, or AAIF.
70 integrations by multiple platforms, approximately 1000000 skills, and nearly 280000000 installations have not yet been supported by original materials in the provided evidence.
The specific governance charter, version management process, and security guidelines for joint maintenance have not been provided.
The evidence provided does not specify 76 the detection scope, total number of samples, or subsequent handling of malicious skills.
Related Developments
Loading event timeline…
Related concepts
AI Agent
that can understand information, reason, and autonomously call tools and perform actions after obtaining the requisite permissions AI system.
