/worktree create feature/auth

~/tmp/worktrees/
├── obsidian-ai-agent/
│   ├── feature-auth/           # branch: feature/auth
│   ├── feature-payments/       # branch: feature/payments
│   └── fix-login-bug/          # branch: fix/login-bug
└── another-project/
    └── feature-dark-mode/

undefined

</core_concepts>

<config>

{
  "terminal": "ghostty",
  "terminalPreference": "auto",
  "enableNumberedTabs": true,
  "shell": "zsh",
  "defaultModel": "opus",
  "claudeCommand": "claude --model opus --dangerously-skip-permissions",
  "portPool": { "start": 8100, "end": 8199 },
  "portsPerWorktree": 2,
  "worktreeBase": "~/tmp/worktrees",
  "defaultCopyDirs": [".claude"],
  "envFilePriority": [".env.local", ".env", ".env.example"],
  "autoCleanupOnMerge": true
}

.claude/
├── CLAUDE.md              # Project instructions (auto-loaded by Claude Code)
├── settings.json          # Hooks and permissions
│   ├── hooks.PostToolUse  # Auto-format on Write|Edit (e.g., "bun run format || true")
│   └── permissions.allow  # Whitelisted bash commands for agents
├── agents/                # Custom subagent definitions
│   ├── build-validator.md # Validates builds pass
│   ├── code-architect.md  # Plans implementation
│   ├── code-simplifier.md # Refactors for clarity
│   └── verify-app.md      # End-to-end verification
└── PROJECT_CONTEXT.md     # Session continuity (project-context-skill)

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Write|Edit",
        "hooks": [{ "type": "command", "command": "bun run format || true" }]
      }
    ]
  }
}

For EACH branch (can run in parallel):

1. SETUP
   PROJECT=$(basename $(git remote get-url origin 2>/dev/null | sed 's/\.git$//') || basename $(pwd))
   REPO_ROOT=$(git rev-parse --show-toplevel)
   BRANCH_SLUG=$(echo "feature/auth" | tr '/' '-')
   WORKTREE_PATH=~/tmp/worktrees/$PROJECT/$BRANCH_SLUG

2. ALLOCATE PORTS
   Find 2 unused ports from 8100-8199, add to registry

3. CREATE WORKTREE
   mkdir -p ~/tmp/worktrees/$PROJECT
   git worktree add $WORKTREE_PATH -b $BRANCH

4. COPY PROJECT CONFIG (.claude/ directory)
   cp -r .claude $WORKTREE_PATH/ 2>/dev/null || true
   Copy .env.local or .env as appropriate

5. CREATE WORKTREE_TASK.md
   Create detailed task file for agent

6. INSTALL DEPENDENCIES
   Detect package manager, run install command

7. VALIDATE (optional)
   Start server, health check, stop

8. REGISTER IN GLOBAL REGISTRY
   Update ~/.claude/worktree-registry.json with entry

9. LAUNCH AGENT
   ghostty -e "cd $WORKTREE_PATH && claude --model opus --dangerously-skip-permissions"

AFTER ALL COMPLETE:
- Report summary table to user

~/.claude/skills/worktree-manager/scripts/status.sh
~/.claude/skills/worktree-manager/scripts/status.sh --project my-project

cat ~/.claude/worktree-registry.json | jq -r '.worktrees[] | "\(.project)\t\(.branch)\t\(.ports | join(","))\t\(.status)"'

~/.claude/skills/worktree-manager/scripts/cleanup.sh <project> <branch> --delete-branch

undefined

**Step 3: Tab Numbering**
The launch-agent.sh script automatically numbers iTerm2 tabs as `[1] project - branch`, `[2] project - branch`, etc. when using `--terminal iterm2`.

undefined

**Workflow:**
1. Start work in terminal
2. When leaving desk, run `/teleport`
3. Continue from https://claude.ai/code
4. Or check in from Claude iOS app
5. Resume locally: `claude --resume <session-id>`

spin up worktree for feature/big-refactor

/ralph-loop "Implement complete auth system with tests" --max-iterations 10

/ralph-loop "Refactor database layer" --completion-promise "All tests pass and no type errors"

/ralph-loop "Implement feature X" --completion-promise "verify-app returns OVERALL: PASS"

undefined

**Parallel Work:**
```bash

**End of Day:**
```bash

undefined