cmux-cli
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill enables the agent to interact with terminal surfaces using commands like
cmux sendandcmux send-key. This allows the execution of arbitrary shell commands within managed panels, which is consistent with the skill's purpose as a terminal manager. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from the terminal environment that could potentially be influenced by untrusted processes.
- Ingestion points: State information and metadata captured via
cmux identify --json,cmux list-panels --json, andcmux sidebar-state. - Boundary markers: The skill encourages the use of structured JSON output (
--json) to separate command metadata from terminal content. - Capability inventory: Full terminal access including the ability to send text and keys to execute commands.
- Sanitization: No explicit sanitization or filtering is performed on the content retrieved from the terminal surfaces or logs before it is presented to the agent.
Audit Metadata