2026-04-07
Gate: all 4 open PRs (#106, #110, #112, #114) merge or get positive signal before starting day 1.
| PR | Branch | Lines | Type |
|---|---|---|---|
| #106 | fix/zoom-state | +210 | Bug fix |
| #110 | fix/headless-rpc-crash | +22 | Bug fix |
| #112 | fix/deserialized-pane-id | +264 | Bug fix |
| #114 | feat/vaxis-text-input | +492 | Behavioral change |
All feedback addressed. All CI green. 6 PRs already merged upstream.
Three small PRs that are hard to disagree with. Review is trivial — the maintainer can merge each in minutes.
| Branch | Lines | Pitch |
|---|---|---|
fix/listptys-oom-leak |
+28 | Memory safety fix. 1 file. Companion to merged #110 — same area of server.zig. Proper errdefer chains for nested allocations in handleListPtys and handleGetServerInfo. |
feat/expose-execute-action |
+46 | Bridge keybind handlers to programmatic access. Pure Lua. Config code and dialog callbacks can fire built-in actions like close_tab without reaching into locals. Was PR #102 (closed by us for cleanup). |
feat/expose-theme-api |
+109 | Read-only access to resolved theme colors. No behavior change. Config code styles custom UI to match the active theme instead of hardcoding hex values. |
Goal: Keep the review queue flowing. Small diffs build velocity and trust.
The big move. Two PRs that tell a story together.
| Branch | Lines | Pitch |
|---|---|---|
feat/detached-session |
+264 | Headless session creation: prise -d -s name. Prepare a session before any terminal connects. Self-contained, clear use case, low controversy. |
feat/session-switch-no-exec |
+703 | In-process session switching. Frame as correctness: exec-based switching drops alt-screen, flashes the shell, and loses state. The in-process approach fixes all three. SessionRestorePlan preflights the target before tearing anything down — strictly safer than current behavior. Was PR #104 (closed by us, not rejected). |
Open detached-session first — smaller, provides context for why session management matters. Follow with session-switch-no-exec same day.
Session switching pitch notes:
- Lead with the bugs it fixes, not the features it enables
- "The current exec path drops alt-screen state and resets the shell" is a bug report
- SessionRestorePlan is a safety improvement — preflight before teardown
- Don't mention agent workflows or external orchestration in the PR description
- Frame generically: "users switching between projects lose their place"
After session-switch-no-exec gets positive signal.
| Branch | Lines | Gate |
|---|---|---|
feat/keep-attached |
+227 | Positive signal on session-switch-no-exec |
feat/nested-session-switch |
+242 | session-switch-no-exec merged or approved |
keep-attached is a config option — small, independent. nested-session-switch depends on session-switch-no-exec and adds RPC-triggered switching.
Ordered by review cost, interleave small and large.
| Branch | Lines | Type |
|---|---|---|
feat/tab-info-api |
+274 | Read-only API |
feat/input-forwarding |
+249 | New capability |
feature/wrap-focus-movement |
+133 | Behavioral change |
feat/tab-bar-state |
+230 | Read-only API |
fix/tab-bar-click-regions |
+204 | Behavioral change |
feat/tab-name-from-cwd |
+186 | Behavioral change |
| Branch | Lines | Type |
|---|---|---|
feat/rename-tab |
+1142 | New capability |
feat/spawn-pty |
+731 | New capability |
fix/stale-session-attach |
+682 | Bug fix (large refactor) |
fix/crash-context |
+628 | Infrastructure |
feat/multi-client-resize-policy |
+35 | New capability (needs design review) |
feat/lua-attach-pty |
+87 | New capability (pairs with spawn-pty) |
| Branch | Lines | Reason |
|---|---|---|
feat/test-filter |
+4 | Rejected (PR #113 closed by maintainer) |
feat/mmap-screen-dump |
+499 | Experiment |
feat/resize-boundary-overlay |
+30 | Experiment (visual approach needs work) |
Too many open PRs. If the current 4 haven't moved by day 2, hold the session switching PRs. Don't flood a stalled queue.
Session switching gets pushback. Detached-session stands alone. Keep-attached and nested-switch wait. The core pitch (exec drops state) is hard to argue with — the approach might get design feedback, but the problem statement is solid.
Maintainer goes quiet. Small PRs don't rot. They sit there as evidence of clean, useful contributions. When the maintainer comes back, there's a buffet of easy merges.
- Small PRs move fast. 28 lines reviews faster than 700. Front-load the easy ones.
- Frame features as fixes when honest. Session switching fixes real bugs (alt-screen, state loss). Lead with that.
- Don't pitch agent infrastructure. Every PR description talks about what prise users get, not what external tooling needs.
- Pace the queue. Watch how fast PRs move. Speed up or slow down accordingly.
- Each PR stands alone. No "this is part 3 of 7" energy. Every PR is independently useful.
cwd: /Users/mike/src/possibilities--prise
session-id: cd826b39-b3cc-4ec1-8a66-17d951387a5f
session-name: possibilities--prise-cd826b39
path: /Users/mike/docs/prise-pr-next-steps.mdcd /Users/mike/src/possibilities--prise && claude --resume cd826b39-b3cc-4ec1-8a66-17d951387a5f