diff --git a/core-maintainers/2025-12-11.md b/core-maintainers/2025-12-11.md new file mode 100644 index 0000000..2c8e9be --- /dev/null +++ b/core-maintainers/2025-12-11.md @@ -0,0 +1,43 @@ +# Core Maintainers Meeting - 2025-12-11 + +## Attendees + +## Agenda + +- **RFDs** + - [Proxy Chains](https://github.com/agentclientprotocol/agent-client-protocol/pull/243) (@nikomatsakis) + - Working prototype being refined for proxy middleware functionality + - Enables two key features: + - Initialize as proxy with message forwarding to client/agent predecessors/successors. Able to compose components together, so you can affect your ACP setup without implementing an entire agent or client + - Extending something like "claude skills" across multiple agents + - "ACP" MCP transport + - Ready for draft merge pending final tweaks, asked for vibe check from the team + - no major concerns raised + - Demo planned for end of week + - [ACP Agent Registry](https://github.com/agentclientprotocol/agent-client-protocol/pull/289) (@ignatov) + - Move from links on GitHub to downloading agents from Registry + - Zed could move to support downloading agents from the registry + - could also tie-in to proxy components, host in additional to agents themselves + - Another goal is to allow people to use a custom registry that works with a common spec + - basically just a json file with a specific format + - would be nice to have this ~Q2 2026 + - Authentication flow for agents [auth](https://github.com/agentclientprotocol/agent-client-protocol/pull/330/files) (@anna239) + - currently, most agents don't supply auth methods + - in general, can we provide agents with the ability to present a gui or request data from the user? + - for JetBrains it would be nice to have a separate flow for authentication + - different UX flow needed + - need to be careful to avoid phishing issues, especially once we're downloading agents from the registry + - Most agents right now are supporting api keys, allow to reuse key from the keychain + - Next Edit Suggestion [NES](https://github.com/agentclientprotocol/agent-client-protocol/pull/325), @ignatov + - Current RFD isn't where we'd want it for an API, but it at least provides an anchor for discussion + - less of a common workflow across edit prediction providers, especially around how they handle context gathering + - some are just a model, you send them a prompt they provide a diff + - others have an API with specific fields they want + - Agus: tradeoff of value protocol brings vs constraints it imposes + - can we get an idea from the providers on what they would expect from an API? +- Update about the governance, @ignatov, @benbrandt + +## Action items + +- [ ] [@all] everyone take a look at the auth RFD and give comments +- [ ] [@ignatov] will gather feedback from providers, @jetbrains + @zed EP teams will give input too