the-gigithe-gigi/learn-claude-code/skills/cmux-close-browser-surfaces/SKILL.md

cmux-close-browser-surfaces

Skill1 starsChanged 6 months agoMITView on GitHub

Bulk-close cmux browser surfaces. Default scope is the caller's current workspace (across all panes); `--workspace <ref>` targets a specific workspace; `--all` covers every workspace in the current cmux window. `--list` previews matches without closing; `--exclude <ref>...` keeps specified surfaces around. Use whenever the user asks to "close the browser surface(s)", "close the cmux browser tab(s)", or otherwise tidy up cmux browser surfaces left behind by prior work. Terminal surfaces are never touched. cmux is a tmux-like multiplexer with terminal and browser surfaces; this skill is unrelated to web-browser tabs or OS-level browser windows.

In this repository3 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- docker-desktopSkill
- generate-imageSkill

The file

Create a free account to mount

Files

SKILL.md

Close cmux Browser Surfaces

Close every browser surface in scope using the cmux CLI. Terminal surfaces are left alone.

When to use

Trigger this skill when the user asks anything like:

"close the browser surface(s) in this workspace"
"close the cmux browser tab(s)"
"kill the browser panes"
"tidy up the browser surfaces"

The cmux CLI is the only supported control plane — do not try to close surfaces by killing processes or manipulating windows directly.

For closing one specific surface, prefer cmux close-surface --surface <ref> directly. This skill exists for the bulk case.

Scope semantics

cmux has windows → workspaces → panes → surfaces. The skill operates on surfaces, never panes or workspaces themselves. Within whatever scope is selected, every pane's surfaces are inspected and only those with type == "browser" are closed.

Flag	What gets considered
(none)	Caller's current workspace, across all its panes.
`--workspace <id\|ref\|index>`	One specific workspace, across all its panes.
`--all`	Every workspace in the current cmux window (does not cross to other windows).

--workspace and --all are mutually exclusive. Cross-window scope is not currently supported — switch to the target window first or invoke per-workspace.

Selection

Flag	Behavior
`--list`	Print browser surfaces in scope and exit. No surfaces are closed. Use this to discover refs before deciding what to close or exclude.
`--exclude <ref>[,<ref>...]`	Skip the listed surface refs. Repeatable; values are unioned. Refs can be UUIDs, `surface:N` refs, or per-workspace indexes — same forms `cmux` itself accepts.

Usage

# Default: caller's workspace, all browser surfaces.
./scripts/close-browser-surfaces.sh

# Specific workspace.
./scripts/close-browser-surfaces.sh --workspace workspace:28

# Every workspace in the current window.
./scripts/close-browser-surfaces.sh --all

# Discover refs first.
./scripts/close-browser-surfaces.sh --list
./scripts/close-browser-surfaces.sh --all --list

# Close all browser surfaces except one.
./scripts/close-browser-surfaces.sh --exclude surface:107

# Close all browser surfaces except two (either form works).
./scripts/close-browser-surfaces.sh --exclude surface:107,surface:113
./scripts/close-browser-surfaces.sh --exclude surface:107 --exclude surface:113

# Close a single specific surface — use cmux directly, not this skill.
cmux close-surface --surface surface:107

Output

Each non-list run prints one line per surface processed:

closed <surface> in <workspace> — <title>\t<url> on success.
skipped <surface> in <workspace> — <title>\t<url> for --exclude'd surfaces.
FAILED to close ... (to stderr) if a cmux close-surface call returns non-zero.

--list prints <workspace>\t<surface>\t<title>\t<url> per row, with a trailing (excluded) column when the surface matches --exclude.

Exit codes: 0 if everything went cleanly (including "no browser surfaces in scope"); non-zero only if a close-surface call actually failed.

Requirements

cmux CLI on PATH.
jq on PATH.
Bash 3.2+ (works on macOS system bash; no mapfile).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free