The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs standard shell commands including sed, cp, test, and date to manage configuration files. It also interacts with a local settings utility located at ~/.agents/skills/cmux-settings/scripts/cmux-settings for listing and validating application preferences. These actions are within the expected scope of a configuration management tool.
[DYNAMIC_EXECUTION]: This skill defines and generates configuration entries that include shell commands (e.g., bun dev, ssh devbox, lazygit) to be executed within the cmux terminal environment. The generation follows user-defined templates or specific requests and is a core feature of the terminal multiplexer's customization.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface by reading configuration data from project-local and global JSON files. If these files originate from an untrusted repository, they could contain instructions intended to mislead the agent. The skill mitigates this risk by requiring the agent to inspect existing configurations and perform validation steps before finalizing changes.
Ingestion points: ~/.config/cmux/cmux.json, .cmux/cmux.json, and .cmux/dock.json files are read into the agent context.
Boundary markers: The agent is instructed to use sed to preview file contents before modification.
Capability inventory: The skill can execute local shell commands and update terminal actions and workspace layouts.
Sanitization: Relies on the cmux-settings validate command and post-configuration verification steps.

cmux-customization