Should you use Pullboard?
Sometimes no. Agent count alone is the wrong test. The useful question is whether complexity × duration makes rediscovery, collision, or unsupported closure more expensive than just using Pullboard.
Skip it
A throwaway task you’ll never reopen — use the simpler tool.
Board it
Anything you’ll revisit, hand off, or need to trust. It’s nearly free, and agents forget.
Run your real work through it
When to skip, when to board.
- SkipA shell one-liner you’ll run once.Nothing outlives the command.
- SkipA question answered in the current chat.The context is still in the window.
- BoardA build that spans several sessions.Every new session starts cold. The board doesn’t.
- BoardTwo agents in the same repo.Claims keep them off each other’s work.
- BoardWork you’ll pick back up tomorrow.You reopen to state, not “where was I?”
- BoardA task someone has to trust is done.Closure is recorded, not claimed.
Why Pullboard wins
Sessions end. The board doesn’t.
Every agent starts blank. The board carries one task from ready to done across every session — so the work moves forward even when the context doesn’t.
Session 1 · MonFresh context — no memory of last week.
Session 2 · TueNew window, new agent — still blank.
Session 3 · WedA different agent entirely.
Checkout flowready
Checkout flowclaimed
Checkout flowdone
PULLBOARD · one board carries the work across every session