You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I wanted to start by saying I've been using squad for a few weeks now and loving it (not the GitHub rate limits for sessions / weekly limits on old annual billed pro plan).
I am running into an issue where agents are occasionally getting lost when starting a new session and fresh context.
There's a TURN_2_SPRINT_PLAN.md with checkboxes and also tracking things via GitLab.
The agents are working on closed GitLab issues because the checkboxes in the plan are not checked.
Seems like it checked all the usual locations to figure out where to resume but couldn't.
When using GitLab we need some way to authoritatively control the work.
My team started working on old GitLab Issue #28 when all open issues were closed. The last one closed was #34.
What should have happened is to evaluate work that needed kicked off, create glab issues, and assign them out to the team to work.
This seems to happen quite a bit that between sessions the team is just lost when starting a new sessions with a fresh context and most of the time figures it out but gets lost other times.
We added squad watch and squad loop in a previous release. Would either of those suffice? I'll look at the GSD done package, though. The name alone... chef's kiss
Activity
bradygaster commentedon May 18, 2026
We added
squad watchandsquad loopin a previous release. Would either of those suffice? I'll look at the GSD done package, though. The name alone... chef's kissfix(coordinator): resolve externalized state, teamRoot, stale work, a…