Anthropic transfers Agent Skills standard to independent foundation

The 70-platform adoption figure is doing a lot of work here. What the summary doesn't surface is that donating a standard after it has already achieved critical mass is a very different act than donating one to seed adoption. Anthropic is ceding formal control of something that has already won, which limits competitive risk while maximizing goodwill.

This fits a pattern visible across Modelwire's recent Anthropic coverage: the company is systematically building governance infrastructure around its products rather than just the products themselves. The METR audit access reported by The Decoder on October 16 showed Anthropic accepting external oversight of its models. The law firm partnerships from October 18 showed it trying to set norms before agents operate autonomously. Donating Agent Skills to a neutral foundation follows the same logic applied to interoperability standards. Taken together, these moves suggest Anthropic is deliberately trading direct control for institutional legitimacy ahead of its IPO, a trade-off that looks rational given the public benefit fund structure announced the same week.

Watch whether OpenAI, which has already adopted the standard, moves to contribute its own agent capability specifications to the foundation within the next two quarters. If it does, the foundation becomes genuinely neutral infrastructure; if it stays a recipient rather than a contributor, Anthropic retains de facto agenda-setting power regardless of the formal governance transfer.