External Requests
PR #7580 (LRX-46412) for the CLF Phase 1 flag removal and PR #7571 (LRX-46413) for the fulfilment popover back navigation fix are both still sitting in my queue. I haven't forgotten about them, they're just queued behind other work, but I want to get to them soon since both are the kind of change that looks trivial in the diff and gets noticed fast if something's off.
I also picked up a new one: PR #7331 for LRX-43659, a teammate's change adding Renovate to webregister for internal packages. Dependency automation tooling like this is worth a careful look, since a misconfigured Renovate setup tends to cause noise for everyone on the team rather than just the person who merged it.
Projects
PR #18320 for LRX-43904, adding connection status pills and banners, is still waiting on my review. On the testing side, LRX-41895 through LRX-41900 remain queued up for validation following last week's bootstrap.ts work, and that testing pass is still ahead of me before LRX-41890 can go out the door.
Dev Environment
I paused the MCP subagent architecture work after validating that OMO MCP lists are permission-only in the current v2 stack. The native v1.19.3 release does filter MCP schemas out of provider requests, which is the behavior I was after, but I decided against migrating to it because doing so would mean losing OMO workflow capabilities that haven't been verified yet. Rather than push forward on an unverified tradeoff, I'm holding here and documented the reasoning in the [deepwork report](file:///Users/jt.houk/.slim/deepwork/opencode-isolated-worker/progress.md).
I've set a clear trigger for picking this back up: once an immutable OpenCode v2 release ships either verified request-level MCP manifest filtering or lazy and deferred MCP schema loading, I'll rerun the hermetic marker fixture before touching the active configuration again. I'm tracking the relevant threads, the OpenCode releases, the lazy-loading proposal, and OMO releases, so I'm not relying on memory to catch the moment this becomes viable again. This is the kind of infrastructure decision that's easy to force through impatiently, but a half-verified migration here would be the kind of thing that only surfaces as a problem weeks later.