Should Anthropic join the Agent Plugins specification (agent-plugins.org)?

[YunosukeYoshino]

Agent Plugins 1.0.0 is an open, vendor-neutral format for packaging Agent Skills and MCP servers into portable plugins. Its initial Technical Steering Committee includes Core Maintainers from Amazon, Cursor, Microsoft, OpenAI, and Vercel. Anthropic is not listed.

That seems like a gap, given that:

• The skills/ component of an Agent Plugin is the Agent Skills format that Anthropic introduced.
• The package layout (plugin.json, skills/, MCP config) is very close to Claude Code's existing plugin format.
• Claude Code is already partially compatible. However, it doesn't select validation rules from the Agent Plugins $schema (anthropics/claude-code#88906), and there are spec conformance gaps such as ${PLUGIN_ROOT} not being expanded (anthropics/claude-code#93250).

As a skill author, I'd like to publish one package that works in Claude Code and other clients without maintaining per-client variants. Having Anthropic take part in the spec, or at least commit to conformance in Claude Code, would make that much more realistic.

Is there any official position on participating in Agent Plugins? Proposals start in its GitHub Discussions, so there's an open path for Anthropic to get involved.