Pure Aliyun ASR skill for voice message transcription, supports multiple channels including Feishu

This skill's code implements Aliyun ASR and calls official Aliyun endpoints, but there are important inconsistencies you should consider before installing: - The README's "no extra configuration" claim is false: you must create /root/.openclaw/aliyun-asr-config.json containing your Aliyun access_key_id/access_key_secret and app_key. The registry metadata did not declare this config path. Confirm you are comfortable storing credentials on disk at that location and that the agent process has permission to read it. - The code invokes ffmpeg for OGG→WAV conversion but ffmpeg is not listed as a required binary. Ensure ffmpeg is available and that calling subprocesses is acceptable in your environment. - The skill posts raw audio bytes to Aliyun NLS endpoints (expected for ASR). There are no hidden external endpoints in the code, which is good, but review the code yourself if you don't fully trust the author. - Prefer creating a least-privilege RAM subuser as recommended, and set strict file permissions (chmod 600) on the config file. Consider running the agent under a non-root account and placing the config in a non-root path — or update the code to allow a configurable config path. If you need this functionality and are comfortable with the above, the implementation is plausible. If you cannot or will not store cloud credentials on disk at /root or cannot allow subprocess calls, do not install. If uncertain, request the author to (1) declare the config path in metadata, (2) allow config path override via env var, and (3) declare ffmpeg as a required binary.

Type: OpenClaw Skill Name: aliyun-asr Version: 1.0.10 The skill is classified as suspicious due to the use of `subprocess.run()` to execute `ffmpeg` in `aliyun_pure_asr.py`. While intended for legitimate audio format conversion, passing a potentially user-controlled `audio_file` path to an external command introduces a shell injection vulnerability risk, even when arguments are provided as a list. This capability, if exploited, could lead to arbitrary command execution within the OpenClaw agent's environment. There is no evidence of intentional malicious behavior like data exfiltration to unauthorized endpoints or prompt injection attempts in SKILL.md.