tmux
Warn
Audited by Gen Agent Trust Hub on May 8, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The 'skill-report.json' file contains self-referential safety claims (e.g., 'risk_level: safe', 'safe_to_publish: true') and attempts to pre-emptively dismiss security findings as 'spurious' or 'false positives'. These metadata claims are designed to deceive security evaluators and bias the analysis verdict.
- [REMOTE_CODE_EXECUTION]: Both 'SKILL.md' and 'skill.json' reference core logic residing in JavaScript/TypeScript files located at '.opencode/skill/tmux/' and 'src/tmux/'. Since these executable files are not present in the provided skill package, the underlying code cannot be audited for malicious logic or vulnerabilities.
- [COMMAND_EXECUTION]: The skill provides an interface for agents to execute arbitrary shell commands in background tmux sessions. This capability allows persistent, unmonitored access to the host environment, which could be exploited to maintain unauthorized access or manipulate background services.
- [PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection via the 'send-command' tool. Untrusted data from user inputs or external sources is interpolated into shell commands without evidence of sanitization, validation, or boundary markers to prevent command injection within the tmux session.
Audit Metadata