byted-voice-to-text

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). The skill accepts and processes audio from arbitrary URLs (scripts/asr.py --url) and downloads Feishu-hosted audio (scripts/asr.py download_feishu_audio), and SKILL.md explicitly instructs treating the script's ASR output as the user's message—so untrusted, user-provided third-party content can be converted into text that the agent will read and act on.