voice-output
Fail
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/speak.shis vulnerable to shell command injection. The variable$TEXTis interpolated directly into system commands likesay,espeak, andfestivalusing double quotes. In a Bash environment, this allows for command substitution (e.g., passing$(id)as text) to execute arbitrary commands on the host system. - [REMOTE_CODE_EXECUTION]: The script
scripts/speak-browser.shuses the Chrome DevTools ProtocolRuntime.evaluateto execute JavaScript in a browser. It injects user-controlled text into a template literal. Because it only escapes backticks and not the${}interpolation syntax, an attacker can execute arbitrary JavaScript within the browser context (e.g., stealing cookies or performing actions on behalf of the user). - [CREDENTIALS_UNSAFE]: The skill includes two undocumented scripts,
scripts/send-email.pyandscripts/send_email.js, which are unrelated to the stated purpose of voice output. These scripts access sensitive environment variables (POPEBOT_EMAIL_USER,POPEBOT_EMAIL_PASS, etc.) and provide a mechanism to send emails via Gmail SMTP. The presence of hidden functionality that handles credentials is a significant security risk. - [EXTERNAL_DOWNLOADS]: The documentation in
SKILL.mdencourages the use ofsudo apt installto download and install external software (espeak), which requires administrative privileges and external network access.
Recommendations
- AI detected serious security threats
Audit Metadata