feishu-voice-sender
Fail
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill accesses sensitive credentials by reading from the local file system at
~/.openclaw/.env. It specifically looks forFEISHU_APP_IDandFEISHU_APP_SECRETto authenticate with the Feishu API. There is also a discrepancy as the documentation suggests a different path (~/.openclaw/config/main.env), which could lead to confusion regarding where secrets are stored. - [COMMAND_EXECUTION]: The script executes the system command
ffmpegviasubprocess.runto handle audio conversion. While it passes arguments as a list, it processes user-supplied file paths directly as inputs to this command. - [DATA_EXFILTRATION]: There is a significant risk of data exfiltration due to a fallback mechanism in the
convert_to_amrfunction. If theffmpegconversion fails (e.g., when a text file or private key is provided instead of audio), the code usesshutil.copyto move the original file to a temporary location and then uploads it to the Feishu servers (https://open.feishu.cn/open-apis/im/v1/files). This behavior allows the agent to be manipulated into uploading arbitrary local files to a remote API.
Recommendations
- AI detected serious security threats
Audit Metadata