Anthropic Skipped Agent Plugins 1.0 — Here's the 3-Step Bridge to Use Them

#ai #tech #opinion #analysis

Agent Plugins 1.0 packages convert to Claude Code with agent-plugins-bridge: relocate plugin.json, rename mcp.json, swap two tokens. Skills work as-is.

What Changed — Agent Plugins 1.0 Launched Without Anthropic

On August 6, 2026, AWS, Microsoft, OpenAI, Vercel, and Cursor shipped Agent Plugins 1.0.0 — a vendor-neutral format for packaging AI agent skills and MCP servers into portable plugins. Google joined as a Core Maintainer the same day. VS Code, GitHub Copilot, Cursor, ChatGPT, and Kiro all support it at launch.

Anthropic isn't on the list. That's ironic because the spec's skills/ component is required to conform to Anthropic's own Agent Skills format — the same one Claude Code has used since its skills system launched.

What It Means For You — The Gap Is Smaller Than You Think

A developer named Yama tested the practical cost of Anthropic's absence. They pulled the real spec repos (agentplugins/agent-plugins-spec and agentplugins/agent-plugins-example), read Claude Code's plugin reference, and found exactly three divergences:

Manifest location: Agent Plugins wants plugin.json at package root. Claude Code wants .claude-plugin/plugin.json. Same fields (name, version, description, author, homepage, repository, license, keywords) — just a different address.
MCP config filename: mcp.json becomes .mcp.json (dot-prefixed). The server definitions inside need zero translation — Claude Code accepts stdio, streamable-http, and sse blocks in the identical shape. Claude Code's docs even list streamable-http as a documented alias for http.
Placeholder tokens: Agent Plugins uses ${PLUGIN_ROOT} and ${PLUGIN_DATA}; Claude Code uses ${CLAUDE_PLUGIN_ROOT} and ${CLAUDE_PLUGIN_DATA}. That's a find-and-replace.

Everything in skills/ needed zero changes. A SKILL.md written for one format is valid for the other — byte for byte.

Try It Now — The Bridge Script

The developer published a single-file Python script called agent-plugins-bridge (github.com/yama3133/agent-plugins-bridge) that does all three transformations with no dependencies beyond the standard library.

# Convert any Agent Plugins package
python3 agent_plugin_to_claude.py agent-plugins-example/ converted-example

# Validate it works in Claude Code
claude plugin validate ./converted-example
# ✔ Validation passed with warnings (just a missing author field)

# Load it into a live session
claude -p --plugin-dir ./converted-example \
  "Do you currently have a skill available named migrate-agent-plugin?"
# Yes — agent-plugins-example:migrate-agent-plugin is available.

The script was tested two ways: against the canonical example package (passed validation) and against a synthetic package with both stdio and streamable-http MCP servers using placeholders in command, args, env, and cwd — all correctly rewritten.

What This Doesn't Settle

The Agent Plugins spec scopes itself to the portable core — skills and MCP servers — leaving distribution, provenance/signing, and permission models to each platform. That's where fragmentation could still happen: five separate marketplaces with five trust models on top of one shared file format.

But for today's practical use: the portable part of the standard works in Claude Code right now, because the underlying skills format was already Anthropic's, and the MCP layer is nearly identical. The entire conversion is renaming two files and swapping two environment variable names.

The Takeaway

Don't wait for Anthropic to join. If you find an Agent Plugins 1.0 package you want in Claude Code, grab the bridge script, run it, and claude plugin validate it. You're one command away from using the ecosystem Anthropic didn't sign onto.

Source: dev.to

Originally published on gentic.news