Fetching from the wire…
Top 5 · 2026-08-13 · source-backed
A spec is a press release until someone who didn't write it implements it.
GitHub made Agent Plugins 1.0 generally available on August 12 across VS Code, Copilot CLI, the Copilot SDK, and the Copilot app on all plans. The spec, published August 6, was co-authored by AWS, Anysphere, Microsoft, OpenAI, Vercel, and Google. A plugin bundles agent skills (in skills/) and MCP servers (mcp.json) into one installable package with a $schema manifest. Vendor-specific behavior lives in namespaced directories like com.github.copilot/. Org governance goes through managed-settings.json with enabledPlugins and strictKnownMarketplaces.
That's a nice spec. Here's the part that makes it real: QwenLM/qwen-code v0.21.11 shipped Agent Plugins v1 support on August 13 (PR #8834). Alibaba is not on the author list. First evidence of adoption outside the authoring group, one week after publication.
Practically, this means a skill-plus-MCP bundle you write once installs across competing CLIs. That's the first time the internal-tooling problem has had a portable answer. Every team I know that has invested in Claude Code skills has quietly accepted they're locked in, and the lock-in isn't the model, it's the packaging. If your skills install into Copilot CLI and qwen-code and Claude Code from the same manifest, model choice becomes a runtime decision instead of an architectural one.
The same qwen-code release has a second idea worth stealing. Its new /coordinate command spawns read-only teammate agents (PR #8804). Read-only is the design choice: parallel agents that can inspect but not write sidestep the write-conflict problem that otherwise forces git worktree isolation for every sibling. Combine that with Anthropic's conformity findings and you get a decent default architecture: many read-only investigators, one writer, explicit merge.
If you maintain internal skills, spend an hour this week converting one to the plugin manifest format. Not because Copilot is where you'll run it, but because that's the format the ecosystem just agreed on, and the cost of conversion goes up with every skill you add.
Each link below shares sources, entities, or timing with this story.
Six clients. One manifest. Zero vendor lock. Vercel published Agent Plugins 1.0.0 on August 6, an openly licensed spec that bundles Agent Skills and MCP servers behind a single portable manifest. The shape is deliberately boring: a plugin.json requiring only schemaVersion and...
Three things happened this month that only make sense together. Agent Plugins 1.0 shipped co-signed by six competitors: AWS, Anysphere, Microsoft, OpenAI, Vercel and Google (GitHub Changelog). It makes skills-plus-MCP bundles portable across clients. OpenAI's August 11 Codex c...
Anthropic invented a file convention. It's now shipping GA inside a competitor's product. Nobody wrote a spec, nobody held a standards meeting, it just happened. On July 29, GitHub made agent skills and MCP server support generally available in Copilot code review for all Pro,...
The sunset notice is short and it doesn't hedge. Code freeze July 29. Repository archived August 10. Core team gone August 31. npm and Docker packages deprecated. Flowise has 55,186 stars and 24,850 forks under Apache-2.0, and the team wrote its own cause of death: developers...
If you wrote an MCP server before July, it's on a protocol shape the maintainers have already removed. Not deprecated-with-a-migration-window. Removed from the spec. MCP lead maintainers David Soria Parra and Den Delimarsky published an updated roadmap on August 22, and the re...
Enterprise-managed MCP allowlists shipped August 6 across the Copilot app, Copilot CLI and VS Code, configured with allowedMcpServers and deniedMcpServers in copilot/managed-settings.json inside the org's .github-private repo. Match by serverUrl with wildcards for remote HTTP/...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.