New Worktree

SkillAI & models

Sets up one or two git worktrees with dependency sync and health-check verification, running parallel setup with two agents when two branches are given. Use when starting parallel feature development in isolated worktrees.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the New Worktree skill

What this skill tells your AI

The instructions your AI receives, as published by coleam00/ai-native-starter-pack in .claude/skills/new-worktrees/SKILL.md and read by ahel’s review.

Quick worktree setup with health check verification. Supports parallel creation with 2 agents.

Parameters

  • Branch 1: $1 (e.g., "feature/search")
  • Branch 2 (optional): $2 (e.g., "feature/export")

Logic

If only $1 provided:

  • Create single worktree sequentially

If both $1 and $2 provided:

  • Spawn 2 agents in parallel using Task tool
  • Each agent sets up their own worktree independently
  • Combine results from both agents
  • Ensure you start the server on a different port to avoid conflicts: worktree 1 dedicated port 8124, worktree 2 dedicated port 8125

Steps

Single Worktree (when only $1 provided)

  1. Create worktree

    git worktree add worktrees/$1 -b $1
    
  2. Navigate to worktree

    cd worktrees/$1
    
  3. Sync dependencies

    uv sync
    
  4. Start server in background

    uv run uvicorn app.main:app --host 0.0.0.0 --port 8124 &
    SERVER_PID=$!
    
  5. Wait for server to be ready

    sleep 3
    
  6. Test health endpoint — make sure you use the correct port

    curl -f http://localhost:8124/health || echo "Health check failed"
    
  7. Kill server

    kill $SERVER_PID
    
  8. Report ready

Parallel Worktrees (when both $1 and $2 provided)

  1. Spawn Agent 1 using Task tool

    Prompt for Agent 1:

    Set up worktree for branch: $1
    
    Steps:
    1. Create worktree: git worktree add worktrees/$1 -b $1
    2. Navigate: cd worktrees/$1
    3. Sync dependencies: uv sync
    4. Start server: uv run uvicorn app.main:app --host 0.0.0.0 --port 8124 &
    5. Wait 3 seconds: sleep 3
    6. Test health: curl -f http://localhost:8124/health
    7. Kill server (find PID and kill it)
    
    Report:
    - Worktree path
    - Branch name
    - Health check result (PASS/FAIL)
    - Any errors encountered
    
  2. Spawn Agent 2 using Task tool (simultaneously with Agent 1)

    Prompt for Agent 2:

    Set up worktree for branch: $2
    
    Steps:
    1. Create worktree: git worktree add worktrees/$2 -b $2
    2. Navigate: cd worktrees/$2
    3. Sync dependencies: uv sync
    4. Start server: uv run uvicorn app.main:app --host 0.0.0.0 --port 8125 &
       (Note: Use port 8125 to avoid conflict with Agent 1)
    5. Wait 3 seconds: sleep 3
    6. Test health: curl -f http://localhost:8125/health
    7. Kill server (find PID and kill it)
    
    Report:
    - Worktree path
    - Branch name
    - Health check result (PASS/FAIL)
    - Any errors encountered
    
  3. Wait for both agents to complete

  4. Combine and report results from both agents

Report

Single Worktree Output:

✓ Worktree initialized
  Path: worktrees/feature-search
  Branch: feature/search

✓ Dependencies synced (uv sync)
✓ Health check passed (http://localhost:8124/health)
✓ Server stopped

Ready for development!

Parallel Worktrees Output:

✓ 2 worktrees initialized in parallel

Agent 1 (feature/search):
  Path: worktrees/feature-search
  Branch: feature/search
  ✓ Dependencies synced
  ✓ Health check passed (port 8124)
  ✓ Server stopped

Agent 2 (feature/export):
  Path: worktrees/feature-export
  Branch: feature/export
  ✓ Dependencies synced
  ✓ Health check passed (port 8125)
  ✓ Server stopped

Both worktrees ready for parallel development!

Notes

  • Exact ports and the server entry point may differ per project — check pyproject.toml and the project README for project-specific values.

Signals

GitHub stars
69
Forks
23
Last commit
Aug 2026
Advanced
Catalog kind
skill
Gateway key
new-worktrees
Source
github.com/coleam00/ai-native-starter-pack