New Worktree
Esta habilidad crea un nuevo _worktree_ de Git para un nombre de rama específico, instalando dependencias y copiando archivos de entorno. Luego abre Visual Studio Code, proporcionando un espacio de trabajo aislado para tareas de desarrollo paralelas.

Qué hace
Esta habilidad simplifica la creación de un nuevo worktree de Git. Configura una nueva rama, instala las dependencias del proyecto (detectando npm, pnpm o bun), copia los archivos .env relevantes y elimina cualquier bloqueo de sesión local de Claude. Finalmente, abre el nuevo worktree en una nueva ventana de VS Code, preparando un entorno de desarrollo aislado para ti.
Cuándo usarlo
Usa esta habilidad cuando necesites trabajar en una nueva característica, una corrección de errores o cualquier tarea que requiera una copia de trabajo de Git aislada sin interferir con tu rama de desarrollo principal. Te permite ejecutar sesiones paralelas de Claude Code, cada una en su propio entorno limpio.
Cómo configurarlo
Para usar esta habilidad, copia su archivo de definición a .claude/skills/new-worktree/SKILL.md en tu repositorio. Luego puedes invocarla dentro de Claude Code escribiendo /new-worktree [prefix/]<name>, por ejemplo, /new-worktree feat/mi-nueva-caracteristica o /new-worktree fix/informe-de-error-123.
El contenido
New Worktree
Spin up a new git worktree for parallel Claude Code session work.
Full reference: @.claude/ref/workflow/worktrees.md
Argument: $ARGUMENTS
Current repo: !git rev-parse --show-toplevel
Current branch: !git branch --show-current
Existing worktrees: !git worktree list
Step 1: Parse and validate input
$ARGUMENTS accepts two forms:
- Bare name — e.g.
discover-bug→ prefix defaults tofeat/ - Prefixed name — e.g.
fix/discover-bug,chore/cleanup-stash,docs/api-readme,refactor/league-adapter
Valid prefixes: feat, fix, chore, docs, refactor, test, hotfix,
ops. Anything else → abort and list the valid set.
The name portion (after the /) must be a non-empty, lowercase,
hyphen-separated identifier. Uppercase letters, spaces, or special characters
other than hyphens → abort and explain.
If $ARGUMENTS is empty → abort and ask the user.
Step 2: Resolve paths
From the current repo path, compute:
repo_root=git rev-parse --show-toplevelrepo_name= basename ofrepo_root(e.g.padel-league.richi.solutions)project_prefix=repo_namewith.richi.solutionsstripped (e.g.padel-league)parent_dir= parent ofrepo_rootprefix= the parsed prefix from Step 1 (defaultfeat)name= the parsed name from Step 1worktree_path=<parent_dir>/<project_prefix>.<name>branch_name=<prefix>/<name>
Examples:
/new-worktree i18n-cleanuponpadel-league.richi.solutions→ worktreepadel-league.i18n-cleanup, branchfeat/i18n-cleanup/new-worktree fix/discover-bugonpadel-league.richi.solutions→ worktreepadel-league.discover-bug, branchfix/discover-bug/new-worktree hotfix/webhook-crashonpadel-league.richi.solutions→ worktreepadel-league.webhook-crash, branchhotfix/webhook-crash
Step 3: Pre-flight checks
Abort with a clear error if:
- The worktree path already exists (
Test-Path/-d) - The branch already exists locally (
git show-ref --verify refs/heads/<branch_name>) - The branch is already checked out in another worktree
(parse
git worktree listoutput)
Step 4: Create the worktree
git fetch origin
git worktree add <worktree_path> -b <branch_name> origin/main
If origin/main does not exist (e.g. solo repo without remote), fall back to
local main.
Step 5: Install dependencies
cd <worktree_path>
npm install
If the repo uses pnpm (detected by pnpm-lock.yaml), use pnpm install
instead. If bun (detected by bun.lockb), use bun install.
Step 6: Copy environment files
For each of these files that exists in repo_root but is in .gitignore,
copy it to worktree_path:
.env.env.local.env.development.env.development.local
Use a non-destructive copy (don't overwrite if it somehow already exists in the new worktree — though Step 3 should have prevented that).
Step 7: Reset session-local state
A fresh worktree must start without inherited Claude session locks. The following files are per-session and should not carry over from another worktree's checkout:
rm -f <worktree_path>/.claude/scheduled_tasks.lock
rm -f <worktree_path>/.claude/*.lock
This is best-effort — missing files are not an error.
Step 8: Open VS Code
code <worktree_path>
This opens a new VS Code window. Tell the user explicitly:
Start a new Claude Code session in the new window. Do not continue the current session — it is attached to the main worktree's workspace and will keep editing files there if you keep using it for code changes.
The current session is fine for reviewing, browsing, or kicking off further worktrees. It is not fine for editing the new feature.
If code is not on PATH, skip the code invocation and tell the user the
absolute path to open manually. Still emit the "new session" guidance.
Step 9: Report
Output a concise summary:
✓ Worktree ready
Path: <worktree_path>
Branch: <branch_name> (based on origin/main)
Deps: installed
Env: copied N file(s)
Locks: cleared
IDE: new VS Code window opened (or: open manually)
➜ Start a NEW Claude Code session in the new window.
Do NOT continue editing in this session — it lives in the main worktree.
Notes
- This skill creates a new branch from
origin/main. It does not check out an existing remote branch. To work on an existing branch, the user should rungit worktree add <path> <branch>manually (no-bflag). - The skill does not commit, push, or merge anything. It only sets up the workspace.
- Cleanup after PR merge is the user's responsibility:
git worktree remove <path>+git branch -d <branch>. A future/cleanup-worktreeskill can automate this.
Viene directamente del framework con el que se construye este sitio. Esta página muestra siempre la versión actual.
¿Quieres algo así para ti?
Empieza con una conversación: 30 minutos, sin compromiso.