toncall-videourl-text

唐潮Toncall工作室开发 - 视频URL语音转文字。支持任意可直接访问的视频URL，自动下载视频、提取音频、上传火山引擎TOS、调用火山引擎语音识别API识别文字，完成后自动清理本地和TOS的临时文件。使用：当用户发送视频URL并要求提取视频文案、语音转文字、转文字时直接执行；当用户只发送视频URL没有说明意...

This skill appears to do what it says: it downloads a provided video URL, extracts audio, uploads audio to your Volcengine TOS bucket, and calls Volcengine ASR to produce text. Before installing or running it: (1) Only use with videos you own or have permission to process — the audio is uploaded to third-party services. (2) Store credentials securely and consider using least-privilege/ephemeral keys for the TOS bucket (upload/delete permissions only for a dedicated bucket/folder). (3) Inspect the full script yourself — the provided file in the package is truncated in the listing and includes minor bugs (e.g., uses datetime.UTC which will fail in standard Python); fix or audit before production use. (4) Run the script in an isolated environment (container) and test with non-sensitive media first. (5) If you want keys kept out of disk, adapt the script to read them from a secure environment or secret store. If you want, I can (a) show the rest of the script if you paste it, (b) list the exact network endpoints and fields sent, or (c) suggest code changes to reduce privileges and improve safety.

Type: OpenClaw Skill Name: toncall-videourl-text Version: 1.0.0 The skill provides video-to-text transcription by downloading videos, extracting audio via ffmpeg, and utilizing Volcengine (ByteDance) TOS and ASR APIs. The Python script `video_url_to_text.py` implements standard AWS V4 signature logic for cloud storage interactions and includes robust cleanup routines for both local and remote temporary files. No evidence of malicious intent, credential exfiltration, or prompt injection was found; the code strictly follows the functionality described in `SKILL.md`.