minhtran3124/agent-harness/using-git-worktrees
- community
Checking that the SKILL.md file is still in the repository.
Set up or verify an isolated worktree before implementation-plan execution. Detect existing isolation first, prefer the harness-native worktree tool, and use a safe git fallback only when needed.
Install
npx vskill@latest install minhtran3124/agent-harness/using-git-worktreesSKILL.md
Using Git Worktrees
Announce that you are setting up an isolated workspace. Run skills/using-git-worktrees/scripts/detect-isolation.sh before creating anything.
linked-worktree: report the path and branch; use it (a detached head needs a branch before
finish).
submodule: do not misclassify it as an isolated worktree; follow its own repository policy.main-worktree: create isolation as below.
Create isolation
Use the native worktree tool if the harness provides one; do not create untracked state the harness cannot manage. In Claude Code, call EnterWorktree with name: "<type>/<kebab-slug>" so the tool itself creates the worktree under .claude/worktrees/. Do not create a checkout by hand and enter it with path: a path outside .claude/worktrees/ raises a "permission-root relocation" confirmation on every entry.
The tool names the branch worktree-<name> (with / written as +), not <type>/<slug>. Rename it with git branch -m <type>/<slug>. Immediately after entering, before any edit, compare git rev-parse HEAD with git rev-parse <base> (the branch the PR will target); if they differ, run git reset --keep <base>, which stops instead of discarding local changes.
Only when no native tool exists, choose an existing .worktrees/ (preferred) or worktrees/ location, or the location documented in CLAUDE.md; if none exists, ask before choosing a root. Ensure the chosen parent is ignored before running:
Use <type>/<kebab-slug> branch names and keep the slug equal to specs/<slug> where possible. Never silently create a nested worktree or allow git add -A to capture one as a gitlink.
Enter the workspace
From inside the selected workspace, deploy the derived harness first:
Then detect project setup from repository manifests and run its clean baseline. Report an existing failure before implementation. writing-plans and subagent-driven-development require this skill; finishing-a-development-branch owns the final PR flow.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/using-git-worktrees/SKILL.md
Popularity
- Repository stars
- 7
- 7-day trend score
- 10
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/minhtran3124/agent-harness/using-git-worktrees)