gh-stack

gh-stack

gh-stack

gh stack is a GitHub CLI extension for managing stacked branches and pull requests. A stack is an ordered list of branches where each branch builds on the one below it, rooted on a trunk branch (typically the repo's default branch). Each branch maps to one PR whose base is the branch below it, so reviewers see only the diff for that layer.

main (trunk)
 └── feat/auth-layer     → PR #1 (base: main)               - bottom (closest to trunk)
  └── feat/api-endpoints → PR #2 (base: feat/auth-layer)
   └── feat/frontend     → PR #3 (base: feat/api-endpoints) - top (furthest from trunk)

The bottom of the stack is the branch closest to the trunk, and the top is the branch furthest from the trunk. Each branch inherits from the one below it. Navigation commands (up, down, top, bottom) follow this model: up moves away from trunk, down moves toward it.

When to use this skill

Use this skill when the user wants to:

• Break a large change into a chain of small, reviewable PRs

• Create, rebase, push, or sync a stack of dependent branches

• Navigate between layers of a branch stack

• View the status of stacked PRs

• Tear down and rebuild a stack to remove, reorder, or rename branches

Prerequisites

The GitHub CLI (gh) v2.0+ must be installed and authenticated. Install the extension with:

gh extension install github/gh-stack

Before using gh stack, configure git to prevent interactive prompts:

git config rerere.enabled true           # remember conflict resolutions (skips prompt on init)
git config remote.pushDefault origin     # if multiple remotes exist (skips remote picker)

Agent rules

All gh stack commands must be run non-interactively. Every command invocation must include the flags and positional arguments needed to avoid prompts, TUIs, and interactive menus. If a command would prompt for input, it will hang indefinitely.

• Always supply branch names as positional arguments to init, add, and checkout. Running these commands without arguments triggers interactive prompts.

• When a prefix is set, pass only the suffix to add. gh stack add auth with prefix feat → feat/auth. Passing feat/auth creates feat/feat/auth.

• Always use --auto with gh stack submit to auto-generate PR titles. Without --auto, submit prompts for a title for each new PR.

• Always use --json with gh stack view. Without --json, the command launches an interactive TUI that cannot be operated by agents. There is no other appropriate flag — always pass --json.

• Use --remote <name> when multiple remotes are configured, or pre-configure git config remote.pushDefault origin. Without this, push, submit, sync, and checkout trigger an interactive remote picker.

• Avoid branches shared across multiple stacks. If a branch belongs to multiple stacks, commands exit with code 6. Check out a non-shared branch first.

• Plan your stack layers by dependency order before writing code. Foundational changes (models, APIs, shared utilities) go in lower branches; dependent changes (UI, consumers) go in higher branches. Think through the dependency chain before running gh stack init.

• Use standard git add and git commit for staging and committing. This gives you full control over which changes go into each branch. The -Am shortcut is available but should not be the default approach—stacked PRs are most effective when each branch contains a deliberate, logical set of changes.

• Navigate down the stack when you need to change a lower layer. If you're working on a frontend branch and realize you need API changes, don't hack around it at the current layer. Navigate to the appropriate branch (gh stack down, gh stack checkout, or gh stack bottom), make and commit the changes there, run gh stack rebase --upstack, then navigate back up to continue.

Never do any of the following — each triggers an interactive prompt or TUI that will hang:

• ❌ gh stack view or gh stack view --short — always use gh stack view --json

• ❌ gh stack submit without --auto — always use gh stack submit --auto

• ❌ gh stack init without branch arguments — always provide branch names

• ❌ gh stack add without a branch name — always provide a branch name

• ❌ gh stack checkout without an argument — always provide a PR number or branch name

• ❌ gh stack checkout <pr-number> when a different local stack already exists on those branches — this triggers an unbypassable conflict resolution prompt; use gh stack unstack first to remove the local stack, then retry the checkout

Thinking about stack structure

Each branch in a stack should represent a discrete, logical unit of work that can be reviewed independently. The changes within a branch should be cohesive—they belong together and make sense as a single PR.

Dependency chain

Stacked branches form a dependency chain: each branch builds on the one below it. This means foundational changes must go in lower (earlier) branches, and code that depends on them goes in higher (later) branches.

Plan your layers before writing code. For example, a full-stack feature might be structured like this (use branch names relevant to your actual task, not these generic ones):

main (trunk)
 └── feat/data-models    ← shared types, database schema
  └── feat/api-endpoints ← API routes that use the models
   └── feat/frontend-ui  ← UI components that call the APIs
    └── feat/integration ← tests that exercise the full stack

This is illustrative — choose branch names and layer boundaries that reflect the specific work you're doing. The key principle is: if code in one layer depends on code in another, the dependency must be in the same branch or a lower one.

Branch naming

Prefer initializing stacks with a prefix (-p). Prefixes group branches under a namespace (e.g., feat/auth, feat/api) and keep branch names clean and consistent. When a prefix is set, pass only the suffix to subsequent add calls — the prefix is applied automatically. Without a prefix, you'll need to pass the full branch name each time.

Staging changes deliberately

The main reason to use git add and git commit directly is to control which changes go into which branch. When you have multiple files in your working tree, you can stage a subset for the current branch, commit them, then create a new branch and stage the rest there:

# You're on feat/data-models with several new files in your working tree.
# Stage only the model files for this branch:
git add internal/models/user.go internal/models/session.go
git commit -m "Add user and session models"

git add db/migrations/001_create_users.sql
git commit -m "Add user table migration"

# Now create a new branch for the API layer and stage the API files there:
gh stack add api-routes # created & switched to feat/api-routes branch
git add internal/api/routes.go internal/api/handlers.go
git commit -m "Add user API routes"

This keeps each branch focused on one concern. Multiple commits per branch are fine — the key is that all commits in a branch relate to the same logical concern, and changes that belong to a different concern go in a different branch.

When to create a new branch

Create a new branch (gh stack add) when you're starting a different concern that depends on what you've built so far. Signs it's time for a new branch:

• You're switching from backend to frontend work

• You're moving from core logic to tests or documentation

• The next set of changes has a different reviewer audience

• The current branch's PR is already large enough to review

One stack, one story

Think of a stack from the reviewer's perspective: the stack of PRs should tell a cohesive story about a feature or project. A reviewer should be able to read the PRs in sequence and understand the progression of changes, with each PR being a small, logical piece of the whole.

When to use a single stack: All the branches are part of the same feature, project, or closely related effort. Even if the work spans multiple concerns (models, API, frontend), they're all building toward the same goal.

When to create a separate stack: The work is unrelated to your current stack — a different feature, a bug fix in an unrelated area, or an independent refactor. Don't mix unrelated work into a single stack just because you happen to be working on both. Start a new stack with gh stack init or switch to an existing stack with gh stack checkout for each distinct effort.

Small, incidental fixes (e.g., fixing a typo you noticed) can go in the current stack if they're trivial. But if a change grows into its own project, it deserves its own stack.

Quick reference

Task Command

Create a stack (recommended) gh stack init -p feat auth

Create a stack without prefix gh stack init auth

Adopt existing branches gh stack init --adopt branch-a branch-b

Set custom trunk gh stack init --base develop branch-a

Add a branch to stack (suffix only if prefix set) gh stack add api-routes

Add branch + stage all + commit gh stack add -Am "message" api-routes

Push branches to remote gh stack push

Push to specific remote gh stack push --remote origin

Push branches + create PRs gh stack submit --auto

Create PRs as drafts gh stack submit --auto --draft

Sync (fetch, rebase, push) gh stack sync

Sync with specific remote gh stack sync --remote origin

Rebase entire stack gh stack rebase

Rebase upstack only gh stack rebase --upstack

Continue after conflict gh stack rebase --continue

Abort rebase gh stack rebase --abort

View stack details (JSON) gh stack view --json

Switch branches up/down in stack gh stack up [n] / gh stack down [n]

Switch to top/bottom branch gh stack top / gh stack bottom

Check out by PR gh stack checkout 42

Check out by branch (local only) gh stack checkout feature-auth

Tear down a stack to restructure it gh stack unstack

Workflows

End-to-end: create a stack from scratch

# 1. Initialize a stack with the first branch
gh stack init -p feat auth
# → creates feat/auth and checks it out

# 2. Write code for the first layer (auth)
cat > auth.go << 'EOF'
package auth

func Middleware(next http.Handler) http.Handler {
    return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
        // verify token
        next.ServeHTTP(w, r)
    })
}
EOF

# 3. Stage and commit using standard git commands
git add auth.go
git commit -m "Add auth middleware"

# You can make multiple commits on the same branch
cat > auth_test.go << 'EOF'
package auth

func TestMiddleware(t *testing.T) {
    // test auth middleware
}
EOF
git add auth_test.go
git commit -m "Add auth middleware tests"

# 4. When you're ready for a new concern, add the next branch
gh stack add api-routes
# → creates feat/api-routes (prefix applied automatically — just pass the suffix)

# 5. Write code for the API layer
cat > api.go << 'EOF'
package api

func RegisterRoutes(mux *http.ServeMux) {
    mux.HandleFunc("/users", handleUsers)
}
EOF
git add api.go
git commit -m "Add API routes"

# 6. Add a third layer for frontend
gh stack add frontend
# → creates feat/frontend (just the suffix — prefix is automatic)

cat > frontend.go << 'EOF'
package frontend

func RenderDashboard(w http.ResponseWriter) {
    // calls the API endpoints from the layer below
}
EOF
git add frontend.go
git commit -m "Add frontend dashboard"

# ── Stack complete: feat/auth → feat/api-routes → feat/frontend ──

# 7. Push everything and create draft PRs
gh stack submit --auto --draft

# 8. Verify the stack
gh stack view --json

Shortcut: If you prefer a faster flow, gh stack add -Am "message" branch-name combines staging, committing, and branch creation into one command. This is useful for single-commit layers but bypasses deliberate staging.

Making mid-stack changes

This is a critical workflow for agents. When you're working on a higher layer and realize you need to change something in a lower layer (e.g., you're building frontend components but need to add an API endpoint), navigate down to the correct branch, make the change there, and rebase.

# You're on feat/frontend but need to add an API endpoint

# 1. Navigate to the API branch
gh stack down
# or: gh stack checkout feat/api-routes

# 2. Make the change where it belongs
cat > users_api.go << 'EOF'
package api

func handleGetUser(w http.ResponseWriter, r *http.Request) {
    // new endpoint the frontend needs
}
EOF
git add users_api.go
git commit -m "Add get-user endpoint"

# 3. Rebase everything above to pick up the change
gh stack rebase --upstack

# 4. Navigate back to where you were working
gh stack top
# or: gh stack checkout feat/frontend

# 5. Continue working — the API changes are now available

Why this matters: If you make API changes on the frontend branch, those changes will end up in the wrong PR. The API PR won't include them, and the frontend PR will have unrelated API diffs mixed in. Always put changes in the branch where they logically belong.

Modify a mid-stack branch and sync

When you need to revisit a branch after the initial creation (e.g., responding to review feedback):

# 1. Navigate to the branch that needs changes
gh stack bottom
# or: gh stack checkout feat/auth
# or: gh stack checkout 42  (by PR number)

# 2. Make changes and commit
cat > auth.go << 'EOF'
package auth
// updated implementation
EOF
git add auth.go
git commit -m "Fix auth token validation"

# 3. Rebase everything above this branch
gh stack rebase --upstack

# 4. Push the updated stack
gh stack push

Routine sync after merges

# Single command: fetch, rebase, push, sync PR state
gh stack sync

Squash-merge recovery

When a PR is squash-merged on GitHub, the original branch's commits no longer exist in the trunk history. gh stack detects this automatically and uses git rebase --onto to correctly replay remaining commits.

# After PR #1 (feat/auth) is squash-merged on GitHub:
gh stack sync
# → fetches latest, detects the merge, fast-forwards trunk
# → rebases feat/api-routes onto updated trunk (skips merged branch)
# → rebases feat/frontend onto feat/api-routes
# → pushes updated branches
# → reports: "Merged: #1"

# Verify the result
gh stack view --json
# → feat/auth shows "isMerged": true, "state": "MERGED"
# → feat/api-routes and feat/frontend show updated heads

If sync hits a conflict during this process, it restores all branches to their pre-rebase state and exits with code 3. See Handle rebase conflicts for the resolution workflow.

Handle rebase conflicts (agent workflow)

# 1. Start the rebase
gh stack rebase

# 2. If exit code 3 (conflict):
#    - Parse stderr for conflict

...