Using git worktrees

Using Git Worktrees

Overview

Ensure work happens in an isolated workspace. Prefer your platform's native worktree tools. Fall back to manual git worktrees only when no native tool is available.

Core principle: Detect existing isolation first. Then use native tools. Then fall back to git. Never fight the harness.

Announce at start: "I'm using the using-git-worktrees skill to set up an isolated workspace."

Step 0: Detect Existing Isolation

Before creating anything, check if you are already in an isolated workspace.

GIT_DIR=$(cd "$(git rev-parse --git-dir)" 2>/dev/null && pwd -P)
GIT_COMMON=$(cd "$(git rev-parse --git-common-dir)" 2>/dev/null && pwd -P)
BRANCH=$(git branch --show-current)

Submodule guard: GIT_DIR != GIT_COMMON is also true inside git submodules. Before concluding "already in a worktree," verify you are not in a submodule:

# If this returns a path, you're in a submodule, not a worktree — treat as normal repo
git rev-parse --show-superproject-working-tree 2>/dev/null

If GIT_DIR != GIT_COMMON (and not a submodule): You are already in a linked worktree. Skip to Step 2 (Project Setup). Do NOT create another worktree.

Report with branch state:

• On a branch: "Already in isolated workspace at <path> on branch <name>."
• Detached HEAD: "Already in isolated workspace at <path> (detached HEAD, externally managed). Branch creation needed at finish time."

If GIT_DIR == GIT_COMMON (or in a submodule): You are in a normal repo checkout. Continue to Step 0b.

Step 0b: Inventory Existing Worktrees and Branches

Step 0 answers "am I isolated?". This step answers "what else is already outstanding?". Skipping it is how a phase-2 worktree gets created beside a phase-1 worktree whose work never landed.

Run Check A of ../using-razorback/references/source-control-hygiene.md:

git worktree list
git -C <each listed path> status --short --branch
git branch --no-merged <base-branch>

For each worktree other than the one you are standing in, and each unmerged branch, classify it with git log --oneline <base>..<branch> (empty result means merged).

Report what you found before creating anything. One line per outstanding item:

Existing worktrees: <path> on <branch> — <N unmerged commits | clean, merged | dirty (<M> files)>

Then decide, and say which you chose:

• Continuation — the new work builds on an unmerged branch. Branch from that branch, or reuse its worktree. A sibling cut from the base cannot see that branch's code and will conflict later.
• Sibling — the new work is genuinely independent. Create the new worktree and state that the outstanding work stays outstanding.
• Reuse — an existing worktree already matches this task. Use it; do not create a second one.

Finding stranded work does not block worktree creation, and it is not a blocker-taxonomy stop. It obliges disclosure. Creating a worktree silently beside unmerged work is the failure; creating one deliberately beside it, and saying so, is fine.

Worktrees outside razorback-managed locations belong to the user. Report them; never remove or reuse them without saying so.

Step 1: Create Isolated Workspace

You have two mechanisms. Try them in this order.

1a. Native Worktree Tools (preferred)

Do you already have a way to create a worktree? It might be a tool with a name like EnterWorktree, WorktreeCreate, a /worktree command, or a --worktree flag. If you do, use it and skip to Step 2.

Native tools handle directory placement, branch creation, and cleanup automatically. Using git worktree add when you have a native tool creates phantom state your harness can't see or manage.

Only proceed to Step 1b if you have no native worktree tool available.

1b. Git Worktree Fallback

Only use this if Step 1a does not apply — you have no native worktree tool available. Create a worktree manually using git.

Directory Selection

Follow this priority order. Explicit user preference always beats observed filesystem state.

1. Check your instructions for a declared worktree directory preference. If the user has already specified one, use it without asking.

2. Check for an existing project-local worktree directory:

   ls -d .worktrees 2>/dev/null     # Preferred (hidden)
   ls -d worktrees 2>/dev/null      # Alternative
   

   If found, use it. If both exist, .worktrees wins.

3. Check for an existing global directory:

   project=$(basename "$(git rev-parse --show-toplevel)")
   ls -d ~/.config/razorback/worktrees/$project 2>/dev/null
   

   If found, use it.

4. Check CLAUDE.md for a stated preference:

   grep -i "worktree.*director" CLAUDE.md 2>/dev/null
   

   If a preference is specified, use it without asking.

5. Default to .worktrees/ at the project root. Razorback runs to completion; do not stop to ask for a directory preference. The user can override the location explicitly in their request.

Safety Verification (project-local directories only)

MUST verify directory is ignored before creating worktree:

git check-ignore -q .worktrees 2>/dev/null || git check-ignore -q worktrees 2>/dev/null

If NOT ignored: Add to .gitignore, commit the change, then proceed.

Why critical: Prevents accidentally committing worktree contents to repository.

Global directories (~/.config/razorback/worktrees/) need no verification — outside project entirely.

Create the Worktree

project=$(basename "$(git rev-parse --show-toplevel)")

# Determine path based on chosen location
# For project-local: path="$LOCATION/$BRANCH_NAME"
# For global: path="~/.config/razorback/worktrees/$project/$BRANCH_NAME"

git worktree add "$path" -b "$BRANCH_NAME"
cd "$path"

Sandbox fallback: If git worktree add fails with a permission error (sandbox denial), tell the user the sandbox blocked worktree creation and you're working in the current directory instead. Then run setup and baseline verification in place.

Step 2: Project Setup

Run the project-documented setup command. Prefer repo docs, existing scripts, CI metadata, or manifest-declared commands. If the repo has no setup guidance, use ecosystem detection as a fallback and record the choice.

<project-defined setup command>

Step 3: Verify Clean Baseline

Run the project-defined baseline or smoke verification scope to ensure the worktree starts clean. Use the target repo's docs as the source of truth.

<project-defined baseline or smoke command>

If baseline verification fails: Treat it as a real blocker unless the plan explicitly says to work from a failing baseline. Report the failure with the command and short output summary. Do not proceed on a broken baseline by default.

If baseline verification passes: Report ready.

If the project has no automated test suite (content-only repos, pure documentation, skill libraries, config-only projects): skip this step. Note in your report that the baseline is manual-verify-only and list the integration checks a reviewer would run instead (schema validation, link checks, manifest validation).

Report

Worktree ready at <full-path>
Baseline verification passing (<N> checks, 0 failures)
Ready to implement <feature-name>

Quick Reference

Situation	Action
Already in linked worktree	Skip creation (Step 0)
In a submodule	Treat as normal repo (Step 0 guard)
Sibling worktree has unmerged commits	Report it, then choose continuation / sibling / reuse (Step 0b)
Sibling worktree is dirty	Report it before creating anything (Step 0b)
New work builds on an unmerged branch	Branch from that branch — don't cut a sibling from the base (Step 0b)
Native worktree tool available	Use it (Step 1a)
No native tool	Git worktree fallback (Step 1b)
.worktrees/ exists	Use it (verify ignored)
worktrees/ exists	Use it (verify ignored)
Both exist	Use .worktrees/
Neither exists	Check CLAUDE.md, then default to .worktrees/
Global path exists	Use it
Directory not ignored	Add to .gitignore + commit
Permission error on create	Sandbox fallback, work in place
Baseline verification fails	Report blocker; do not proceed unless plan allows it

Common Mistakes

Fighting the harness

• Problem: Using git worktree add when the platform already provides isolation
• Fix: Step 0 detects existing isolation. Step 1a defers to native tools.

Skipping detection

• Problem: Creating a nested worktree inside an existing one
• Fix: Always run Step 0 before creating anything

Accumulating stranded worktrees

• Problem: Phase 2 gets a fresh worktree while phase 1's worktree still holds unmerged commits. Nobody notices until the run reports "complete" with work stranded in two places.
• Fix: Step 0b inventories and reports every outstanding worktree and branch before creation, and prefers continuation over a sibling cut from the base

Skipping ignore verification

• Problem: Worktree contents get tracked, pollute git status
• Fix: Always use git check-ignore before creating project-local worktree

Assuming directory location

• Problem: Creates inconsistency, violates project conventions
• Fix: Follow priority: existing > global > CLAUDE.md > default

Proceeding with failing baseline

• Problem: Can't distinguish new bugs from pre-existing issues
• Fix: Report the baseline as blocked unless the plan explicitly allows a failing baseline

Red Flags

Never:

• Create a worktree without running the Step 0b inventory first
• Create a worktree silently beside another worktree that holds unmerged or uncommitted work
• Cut a sibling worktree from the base when the new work builds on an unmerged branch
• Remove or reuse a worktree outside a razorback-managed location without saying so
• Create a worktree when Step 0 detects existing isolation
• Use git worktree add when you have a native worktree tool (e.g., EnterWorktree). This is the #1 mistake — if you have it, use it.
• Skip Step 1a by jumping straight to Step 1b's git commands
• Create worktree without verifying it's ignored (project-local)
• Skip baseline verification
• Proceed with failing baseline unless the plan explicitly allows it
• Stop to ask for a worktree directory preference (default to .worktrees/)

Always:

• Run Step 0 detection first, then the Step 0b inventory
• State what the inventory found and which choice you made (continuation / sibling / reuse)
• Prefer native tools over git fallback
• Follow directory priority: existing > global > CLAUDE.md > default
• Verify directory is ignored for project-local
• Auto-detect and run project setup
• Verify clean baseline

Integration

Called by:

• brainstorming (Phase 4) - REQUIRED when design is approved and implementation follows
• subagent-driven-development - REQUIRED before executing any tasks
• executing-plans - REQUIRED before executing any tasks
• Any skill needing isolated workspace

Escape hatch: The "REQUIRED" callers above default to worktree isolation. The user can opt out with explicit consent for small, same-session work where a feature branch on the current workspace is sufficient. Record the consent and proceed — don't silently skip worktree setup.

Never called by: razorback:fixing-small-issues. Quick-fix-tier work happens on the current checkout by policy — no worktree, no baseline verification, and no consent question needed, because this skill is never invoked for that tier.

Pairs with:

• finishing-a-development-branch - REQUIRED for cleanup after work complete