tmux
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
tmuxcommands extensively to manipulate terminal sessions, including creating, killing, and renaming sessions, as well as sending raw keystrokes and scraping pane content. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from terminal outputs (via
tmux capture-paneand thewait-for-text.shscript). This creates an attack surface where an interactive process running inside a tmux pane could output malicious instructions intended to influence the agent's next actions. - Ingestion points: Pane output is captured in
SKILL.mdexamples andscripts/wait-for-text.shusingtmux capture-pane. - Boundary markers: None identified; the skill processes raw terminal output.
- Capability inventory: Subprocess execution through
tmux, shell script execution, and interaction with other interactive processes. - Sanitization: None identified; output is piped directly to
greportailand returned to the agent context.
Audit Metadata