skill-system-tmux
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various system commands including tmux, ps, lsof, and kill to manage session lifecycles and resource cleanup as described in SKILL.md.
- [DATA_EXFILTRATION]: Through the capture-pane operation, the skill can read terminal output. This provides a mechanism for the agent to ingest data displayed within the terminal environment, which may contain sensitive information depending on the task being performed.
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface (Category 8).
- Ingestion points: Terminal output is read into the agent context via the tmux capture-pane -p -t {session} command in SKILL.md.
- Boundary markers: The skill suggests using shell prompt patterns (e.g., regex: /(\s*|>|#)\s*$/m) to detect command completion but lacks explicit security boundaries to prevent the agent from executing instructions potentially embedded in terminal output.
- Capability inventory: All skill operations (route-command, manage-session, reclaim-resources) utilize proc.exec capabilities through shell execution tools as defined in SKILL.spec.yaml.
- Sanitization: There are no instructions for sanitizing or escaping the content captured from the terminal before it is processed by the agent.
Audit Metadata