safe-editing

Ensure AI agents work in an isolated Git worktree to prevent changes to the main working directory. Use when AI is about to make its first code modification in a session, or when the user requests isolated/safe editing. Triggers include starting to edit files, implementing features, or fixing bugs.

⭐6stars🔀0forks📁View on GitHub🕐Updated Jan 4, 2026⬇️1downloads

When & Why to Use This Skill

The Safe Editing skill provides a robust safety layer for AI-driven development by isolating code changes within dedicated Git worktrees. It ensures that AI agents can implement features, fix bugs, or refactor code without polluting the main working directory or interfering with uncommitted local changes, thereby maintaining repository integrity and streamlining the peer-review process.

Use Cases

  • Feature Development: Automatically spawning an isolated environment for new feature implementation to keep the primary workspace clean and stable.
  • Safe Bug Fixing: Resolving critical issues in a separate worktree to prevent accidental overwrites or conflicts with ongoing manual development work.
  • Risk-Free Code Exploration: Enabling AI agents to perform experimental refactoring or complex architectural changes in a sandbox-like Git state before merging.
namesafe-editing
descriptionEnsure AI agents work in an isolated Git worktree to prevent changes to the main working directory. Use when AI is about to make its first code modification in a session, or when the user requests isolated/safe editing. Triggers include starting to edit files, implementing features, or fixing bugs.

Safe Editing

Use Git worktree to isolate AI-driven changes from the main working directory.

Trigger Conditions

Activate this skill when:

  • AI is about to make its first code modification in the current session
  • User explicitly requests safe/isolated editing
  • Starting a new feature implementation or bug fix

Workflow

1. Check Current State

git rev-parse --show-toplevel
git worktree list
git status --porcelain

Verify:

  • In a Git repository
  • No uncommitted changes in main directory
  • Not already in a worktree

If uncommitted changes exist, ask user how to proceed (stash, commit, or abort).

2. Determine Branch Name

Analyze the task and generate a branch name:

Task Type Branch Format Example
New feature feature/<scope>-<summary> feature/auth-add-oauth
Bug fix fix/<scope>-<summary> fix/api-null-pointer
Refactor refactor/<scope>-<summary> refactor/utils-cleanup
Docs docs/<summary> docs/update-readme
  • Use lowercase kebab-case
  • Translate Japanese to English
  • Keep concise but descriptive

3. Create Worktree

Use the gwt alias (defined in .zshrc):

gwt <branch-name>

This creates:

  • New branch: <branch-name>
  • Worktree at: ../git-worktrees/<repo-name>-<branch-name>

4. Navigate to Worktree

cd ../git-worktrees/<repo-name>-<branch-name>

Confirm the switch:

pwd
git branch --show-current

5. Perform Changes

Execute the requested modifications in the worktree directory. All file edits, creations, and deletions happen here.

6. Commit, Push, and Create PR

Use the commit-push-pr skill to:

  1. Stage and commit changes
  2. Push to remote
  3. Create pull request

7. Cleanup (After PR Merge)

Once the PR is merged, clean up the worktree:

cd <original-repo-path>
gwtr ../git-worktrees/<repo-name>-<branch-name>

Quick Reference

Alias Command Purpose
gwt <branch> git worktree add ../git-worktrees/<repo>-<branch> -b <branch> Create worktree
gwtl git worktree list List worktrees
gwtr <path> git worktree remove <path> Remove worktree

Notes

  • Always confirm with user before creating worktree
  • If worktree already exists for the task, reuse it
  • Keep main directory clean for other work