Agent Plugins 1.0 Defines a Portable Format for Skills and MCP Servers
Agent Plugins introduced its version 1.0.0 working-draft specification on August 6 as a vendor-neutral format for packaging reusable AI-agent components. The format requires a plugin.json manifest and can include Agent Skills and MCP server configuration, while installation, permissions, distribution, and client-specific behavior remain under each compatible product's control.
Agent Plugins introduced a vendor-neutral working draft on August 6 for packaging reusable components that extend AI agents. Version 1.0.0 defines a common directory structure built around a required plugin.json manifest, with optional Agent Skills and MCP server configuration in fixed locations.
The project is not a new agent runtime or permission system. Its stated goal is narrower: give plugin authors a predictable portability layer so the same core components do not need to be rearranged for every compatible client. Distribution, installation, permissions, user experience, and client-specific capabilities remain the responsibility of each product.
What the specification standardizes
The working draft defines two portable component types: Agent Skills and MCP servers. Skills continue to use the external Agent Skills specification, while mcp.json describes stdio, Streamable HTTP, or legacy HTTP+SSE servers. A required manifest identifies the plugin and the Agent Plugins version it targets.
The specification also sets path-containment and loading rules. Plugin-relative paths must remain inside the plugin root, clients must reject or skip invalid components at the narrowest applicable boundary, and portable MCP configuration cannot embed credentials as a standardized secret mechanism. Those requirements make package validation part of interoperability rather than leaving every client to infer its own file layout.
Portability has a deliberate limit
The official project site says the standard is openly licensed and developed in public. Its initial Technical Steering Committee includes core maintainers from Amazon, Cursor, Microsoft, OpenAI, and Vercel, with proposals and material changes handled publicly through GitHub discussions.
That multi-vendor governance matters because the format is useful only if independently developed clients interpret the portable core consistently. It does not mean every client implements every component or extension. Reverse-domain namespaces allow products to add client-specific behavior without changing the shared package format.
For teams building reusable agent tooling, the practical test is therefore two-part: validate the portable manifest, Skills, and MCP configuration against the specification, then separately verify each target client's installation flow, supported transports, permissions, and failure behavior. Version 1.0.0 supplies an interoperability floor, not a promise that one package will behave identically everywhere.
Key Points
- 1Agent Plugins 1.0.0 is a working-draft, vendor-neutral package format for reusable AI-agent components.
- 2The portable core standardizes a plugin.json manifest plus optional Agent Skills and MCP server configuration in fixed locations.
- 3Clients still control distribution, installation, permissions, supported components, and product-specific behavior.
Scoring Rationale
The specification addresses a real portability problem for teams distributing Skills and MCP-backed agent tooling across products. Its impact depends on client adoption and conformance because version 1.0.0 is still a working draft and leaves permissions, installation, and product behavior outside the portable core.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
