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
Rank sessions by a composite priority score so the user always knows where their attention has the highest ROI — no more guessing which terminal to check next.
Problem
When you have multiple agent sessions running, you have to manually decide which one to check next. You guess. You pick the one you remember. You often pick wrong, wasting time on a session that's fine while another one has been stuck waiting for you.
The cost of picking wrong compounds: every minute an agent waits for your input is a minute of lost parallel productivity. And when you finally get to the stuck session, you've lost context and need to re-orient.
Proposed Solution
Rank sessions by a composite priority score that accounts for:
Context Depth
How much you've invested in this session (turns, time spent, decisions made). Sessions where you're deepest should rank higher — re-entry cost is lowest and sunk cost of abandoning is highest. (Related: #20)
Project Urgency
User-defined or inferred priority of the project/repo. A production hotfix outranks a refactor.
Wait Time
How long the session has been waiting for human input. Longer wait = higher urgency, weighted against other factors. (Related: #16 bottleneck detection)
Agent State
Is the agent blocked on an approval? Thinking? Actively working? Blocked agents get boosted. (Related: #19 approval queue)
The score should surface a clear answer to: "Where does my attention have the highest ROI right now?"
Acceptance Criteria
Sessions are sorted by priority score, not by recency or alphabetical order
The top session is always the one where user attention creates the most value
Score updates in real-time as agent states change
User can understand why a session is ranked where it is (tooltip or subtle indicator)
Open Questions
Should users be able to manually override/pin priority?
Should project-level priority be explicit (user sets it) or inferred (from git branch names, repo metadata)?
How to handle the cold start when a session is new and has no history?
Summary
Priority: P0 | Category: Attention Routing
Rank sessions by a composite priority score so the user always knows where their attention has the highest ROI — no more guessing which terminal to check next.
Problem
When you have multiple agent sessions running, you have to manually decide which one to check next. You guess. You pick the one you remember. You often pick wrong, wasting time on a session that's fine while another one has been stuck waiting for you.
The cost of picking wrong compounds: every minute an agent waits for your input is a minute of lost parallel productivity. And when you finally get to the stuck session, you've lost context and need to re-orient.
Proposed Solution
Rank sessions by a composite priority score that accounts for:
Context Depth
How much you've invested in this session (turns, time spent, decisions made). Sessions where you're deepest should rank higher — re-entry cost is lowest and sunk cost of abandoning is highest. (Related: #20)
Project Urgency
User-defined or inferred priority of the project/repo. A production hotfix outranks a refactor.
Wait Time
How long the session has been waiting for human input. Longer wait = higher urgency, weighted against other factors. (Related: #16 bottleneck detection)
Agent State
Is the agent blocked on an approval? Thinking? Actively working? Blocked agents get boosted. (Related: #19 approval queue)
The score should surface a clear answer to: "Where does my attention have the highest ROI right now?"
Acceptance Criteria
Open Questions