Context
The previous issues #630, #631, #632, and #633 came from local worktree experiments and should be treated as one consolidated roadmap with core libtmux API work.
This issue is the canonical tracker for Agentic/DX improvements across:
- core API contracts
- async support direction
- concurrency/state semantics
- engine/backend architecture
Goal
Strengthen API contracts and machine-readable behavior for agent-safe automation without fragmenting the project into parallel tracks.
Low-lift ideas
- Publish explicit stable API boundary and deprecation schedule (safe public imports and behaviors).
- Add a concise exception/error contract table for command failures.
- Publish sync-vs-async parity notes (supported / planned / out-of-scope).
- Define/document engine selection semantics and baseline capability fields.
Medium-lift ideas
- Add contract tests for public behavior and tmux-version compatibility.
- Add async cancellation/timeout/backpressure semantics and tests.
- Clarify concurrency/state refresh guarantees for pane/window/session objects.
- Add shared response/event parsing contracts across subprocess/control-mode backends.
High-lift ideas
- Publish machine-readable capability metadata (tmux version matrix + backend capabilities).
- Add versioned structured contracts for machine output/diagnostics.
- Add inspect/plan-style, non-mutating orchestration preview interfaces.
Done when
- Public API and compatibility expectations are explicit.
- Async + concurrency + backend behaviors are documented and regression-tested.
- Automation clients can rely on stable, machine-readable contracts.
Consolidation note
Supersedes and consolidates: