小野语音系统

• Analyzed: May 12, 2026, 4:28 AM
• Findings: 1
• Version: 1.0.0
• Hash: 7f3ff7e4cc5234bfb0ff69aa2cbc5277fc6d5450775c88cfd420da90915357f1
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill utilizes `subprocess.run` to execute local commands like `say` (macOS TTS) and `ffmpeg` (audio conversion), which are necessary for its core functionality. While arguments are passed as lists to mitigate simple shell injection, any direct execution of external commands introduces a vulnerability surface. More critically, the `SKILL.md` and `README.md` documentation explicitly demonstrate a 'Telegram integration' example using `subprocess.run(["telegram-send", "--file", audio_file])`. This showcases a capability to send files to an external service via a shell command, which represents a significant risky capability that could be exploited for data exfiltration if the AI agent is compromised through prompt injection, even if the skill itself does not exhibit malicious intent.