remote-edit

name	remote-edit
description	Hand a file to the user's running editor and block until they finish editing it. Use whenever a draft (PR body, ticket, commit message, any generated file) needs the user's review before you act on it, or when a task says "open this in my editor/vim/nvim".
allowed-tools	Bash(bash ${CLAUDE_SKILL_DIR}/remote-edit.sh *)

Live servers (address + cwd):

!`bash /Users/dominicmonroe/.claude/skills/remote-edit/remote-edit.sh list`

Open a file for review

bash ${CLAUDE_SKILL_DIR}/remote-edit.sh <file> [server]

Opens <file> in a new tab and blocks. Omit server when exactly one is live.

Which server

• One live → use it, don't ask.
• Several → pick the one whose cwd contains the file, or is the repo/worktree the task is about. Worktrees are separate servers: …/repo/worktrees/foo is not …/repo — match the exact path the work is in.
• Still ambiguous (same cwd, or file outside every cwd) → ask the user.
• None → tell the user; do not start an nvim.

The list above is a snapshot. If the script exits non-zero mid-session the server died — re-run remote-edit.sh list for a fresh one.

Waiting

Done = the call returns, which happens when the user closes the tab (:wq), not when they save — a save leaves the wait live and the user still editing. It blocks past the foreground timeout into the background; that is normal, keep waiting for completion. A file-modified notification, or a killed/stopped/interrupted task status, is not the finish. Only re-read the file once the call genuinely returns.