Paul’s Perspective:
This matters because businesses investing in AI workflows do not want to rebuild the same capabilities for every assistant, IDE, or coding agent they adopt. A workable cross-platform standard could lower integration costs, improve flexibility, and reduce vendor lock-in, but leaders should also watch adoption closely before standardizing internal development around it.
Key Points in Video:
- The standard packages a plugin as a directory containing a skills folder, a plugin.json manifest, and an mcp.json file.
- Portable elements currently include skills and MCP servers, while commands, hooks, agents, rules, and LSP servers remain client-specific.
- Support is emerging from major players including OpenAI, Google, AWS, and Vercel, which gives the format early credibility.
- Anthropic is a notable exception so far, creating a risk that the market could split into more than one plugin format.
- Local installation is still awkward in some tools because the spec defines structure, but not how plugins are distributed or loaded.
Strategic Actions:
- Define the plugin as a simple folder containing skills, plugin.json, and mcp.json.
- Separate what is portable from what remains client-specific, including commands, hooks, and rules.
- Test migration by converting an existing Anthropic knowledge-work plugin to the new standard.
- Run the migrated skill in a different project to validate cross-tool portability.
- Evaluate installation and distribution limitations, especially where tools still expect marketplace delivery.
- Assess long-term viability based on broader industry adoption and whether absent vendors join the standard.
The Bottom Line:
- A new open standard aims to make AI agent plugins portable across tools like ChatGPT, Codex, Cursor, GitHub Copilot, VS Code, and others by packaging skills and MCP servers in one simple folder.
- The bigger takeaway for businesses is the potential to reduce rework across AI platforms, even though distribution, installation, and some client-specific features still remain fragmented.
Dive deeper > Source Video:
Ready to Explore More?
If your team is weighing how to build reusable AI workflows without boxing yourselves into one platform, we can help map out a practical approach. We work together with clients to sort through the standards, tooling, and implementation tradeoffs in plain business terms.
