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
{{ message }}
Repository navigation
Repair tools for Crew lifecycle edge cases after a restart #346
There is no restart recovery for a Crew's lifecycle (Jackson's principle, #327): if the server stops partway through a Captain's unarchive, some seats are back, the Crew is still marked retired, and the rest of the seats are archived. Today the only repair is for the person to archive the Captain again and then unarchive it (#315 documents this). Agents can't repair it themselves: the thread organize tool can unarchive a seat thread, but nothing can un-retire the Crew.
Idea (Jackson, on #315, 2026-09-26; noted, not decided)
Give the person and agents tools to fix edge cases instead of having the platform recover on its own. For example, an agent tool a Captain can use to un-retire a Crew it commands, so it can finish a restore that stopped partway.
Closing as not planned. With #315, #313, and #352 merged, the edge cases this was filed for are covered well enough without a general repair tool:
A Captain's unarchive stops partway. There's still no automatic recovery, which was Jackson's ruling. The repair is documented in crews.md: archive the Captain again and unarchive it, or click Undo. That re-runs the restore and skips seats that are already back. That's the "repair beats edge-case machinery" answer from Embed the 2026-09-26 review principles in product, process and agent docs #327, so I'm not adding an agent tool for it unless it actually happens in use.
A reserved seat after a restart mid-launch. The launch reporter's startup sweep reports every seat without a thread as never created, so the Captain hears about it. There's one narrow gap left: the leftover reservation can stay on the roster and block asking for that seat again. I filed that as fix(crews): a seat reserved before a restart mid-launch stays on the roster with no thread #376 with a small fix.
If a half-restored Crew shows up in real use, I'll reopen this with that case.
Problem
There is no restart recovery for a Crew's lifecycle (Jackson's principle, #327): if the server stops partway through a Captain's unarchive, some seats are back, the Crew is still marked retired, and the rest of the seats are archived. Today the only repair is for the person to archive the Captain again and then unarchive it (#315 documents this). Agents can't repair it themselves: the thread organize tool can unarchive a seat thread, but nothing can un-retire the Crew.
Idea (Jackson, on #315, 2026-09-26; noted, not decided)
Give the person and agents tools to fix edge cases instead of having the platform recover on its own. For example, an agent tool a Captain can use to un-retire a Crew it commands, so it can finish a restore that stopped partway.
Open questions
archive_crew.Not needed for #315. Filed so the idea isn't lost.