👋 I’ve been thinking about how we document and talk about plugins and skills, and that led to a broader question about whether MCPs and hooks should also live under a shared “Extensions” concept. I think that answer is yes but also wanted to get some feedback on the topic. Claude for example groups Extensions together with claude.md, skills, MCP, subagents, agent teams, hooks, and plugins. So 'Extensions' matches if we adopt it as well and we could even include agent.md, subagents, and agent teams here. Then going a bit deeper on some documentation explanations the rough framing would be: • Skills are reusable capabilities or behaviors • Plugins may package or expose broader functionality, including skills, hooks, and MCP-related integrations • MCP servers extend what the system can access or do in a conversation • Hooks may provide event-based or fine-grained control over conversation flow and behavior From a docs standpoint, I think that could help with: • making the OpenHands paradigm easier to understand quickly • giving a clearer place to put new functionality or where to look for it I don’t think all the details are perfectly clear for me yet but the overall grouping feels directionally right to and potentially easier to communicate once "Extensions" have equal implementation across cli/gui/tui/cloud ➡️ Curious whether this “Extensions” framing feels useful to others from a documentation and OSS onboarding perspective or if you have any thoughts/feedback

It feels about right to me. I believe it was the intention when we made extensions/ Hooks are indeed ways of extending OH, conceptually. Arguably MCP too. Extension = behavior customization, behavior addition, I'd say

If one wants to distribute file based sub-agents (and I do) then that also falls under plugins list in claude code.

Yep, I think this sounds right