@_zidkim working through framing and positioning for the product. went through yc's app and had claude come up with a bunch of questions that are intended to force clear thinking and expose gaps
it gets real interesting. stale context gets updated by a subagent that the main agent calls autonomously or that you could explicitly call to update something.
agents stepping on each other is still tricky*.
right now, it still takes a bit of manual curation and is best either having subfolders per person (tech leads own `./engineering/`, pm owns `./product/` etc.) or having someone take point on keeping things clean.
all changes are logged in a changes.jsonl file so there's explicit paper trail and you can have an agent manage overwrites/bad merge conflicts much the same way you could for code.
*but i'm also working on something new (and of course open source) that will make this much more seamless.
i made a coding agent plugin (works across claude code, codex, cursor) that helps you share and collaborate on your team's shared context layer.
connect context across sessions via your local file system and share/collaborate on those files with your team.
github below.
i made a coding agent plugin (works across claude code, codex, cursor) that helps you share and collaborate on your team's shared context layer.
connect context across sessions via your local file system and share/collaborate on those files with your team.
github below.
i made a coding agent plugin (works across claude code, codex, cursor) that helps you share and collaborate on your team's shared context layer.
connect context across sessions via your local file system and share/collaborate on those files with your team.
github below.