Introduce a Bubble Tea control deck with eligible repositories preselected, keyboard navigation, responsive paging, and detailed explanations for blocked worktrees. Require confirmation before synchronization, allow refreshing the workspace, and retain each repository's latest hub output without changing the existing CLI and JSON workflows.
3.8 KiB
Stack Sync
stack-sync safely runs hub sync across every Git repository in a workspace. It is designed for directory trees such as JezzWTF where the root may not be a repository and a parent repository may contain deliberately untracked, nested repositories.
It provides both a full-screen Bubble Tea control deck for interactive work and a conventional CLI/JSON interface for SSH sessions and automation. Both interfaces use the same discovery and safety engine.
Safety model
Before a repository is eligible, Stack Sync verifies that it:
- is on a branch (not a detached
HEAD); - has at least one remote;
- has no merge, rebase, cherry-pick, revert, or bisect in progress;
- has no modified, staged, deleted, conflicted, or untracked files.
Untracked paths that are themselves discovered nested Git repositories are excluded from the parent repository's dirty check. All other untracked files still block it.
Stack Sync never runs git stash, git reset, git clean, git commit, or any other command that saves or discards work. A repository is checked again immediately before hub sync to reduce the chance of a change slipping in between the plan and apply phases. Blocked repositories are skipped; use --strict to abort the whole run if even one is blocked.
Stack Sync intentionally preserves hub sync semantics. That means hub may delete a local branch when its upstream branch has been deleted and it considers the local branch merged. It warns instead when it finds unpushed or apparently unmerged commits. The interactive confirmation calls this out; review hub help sync before using --yes in automation.
Build and install
Requires Go 1.24 or newer, Git, and hub.
cd stack-sync
go test ./...
go build -o stack-sync .
install -Dm755 stack-sync ~/.local/bin/stack-sync
Usage
Interactive TUI
Launch the control deck for the workspace:
stack-sync tui --root ~/Coding/jwtf
Every eligible repository starts selected. Move with the arrow keys or j/k, toggle the focused repository with Space, select all with a, clear the selection with n, and refresh with r. Press s or Enter to review the destructive-branch warning, then y to begin. Blocked repositories cannot be selected and show their first dirty paths in the detail panel. After a run, focusing a repository shows its latest hub output.
CLI and automation
Scan the current workspace. This is read-only and is the default command:
stack-sync scan --root ~/Coding/jwtf
# equivalent:
stack-sync --root ~/Coding/jwtf
Review the same plan, confirm it, and sync every eligible repository:
stack-sync sync --root ~/Coding/jwtf
For automation, suppress the prompt and optionally require the entire workspace to be clean:
stack-sync sync --root ~/Coding/jwtf --yes --strict
Useful options:
--jobs 4 maximum concurrent inspections or syncs
--timeout 5m per-repository hub sync timeout
--exclude temp skip a directory name anywhere in the tree
--exclude Org/old skip a root-relative path
--json emit structured output
Dependency caches, build outputs, and tool-managed directories such as node_modules, target, .claude, and .codex are skipped during discovery by default. These exclusions only affect repository discovery; they never make real changes inside a discovered repository disappear from its dirty-worktree check.
Exit codes are 0 for a successful scan or sync (planned dirty repositories may be safely skipped), 1 for an operational/sync failure, 2 for invalid or unconfirmed non-interactive use, and 3 when strict mode refuses the plan or a previously eligible repository fails its immediate pre-sync safety check.