Fan out agent workspaces on a remote Mac's cmux from Windows — create a cmux workspace with N split panes by running one `ssh <host> cmux <verb>` per operation, launch agent REPLs (claude, codex, any CLI) in each pane, wait for readiness, and return a pane map for downstream skills. Use when the user wants to build an agent grid on the Mac's cmux, "fan out N agents on the mac", spin up remote cmux workspaces/panes, or inspect (status) and tear down grids built this way. Requires PowerShell 7.1+, key-auth SSH to the Mac, cmux running there, and the one-time socket-gate step (see setup.md). NOT for injecting into or reading existing panes conversationally (future cmux-connector), local Windows pane grids (use psmux-fanout), or driving cmux's embedded browser.
92%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
The workflow reads terminal output from remote Mac panes via `read-screen` to check readiness patterns, which can ingest outsider-authored terminal output if untrusted agents run in those panes.
KennethJefferson/cmux_test
a4e208e
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.