Whisper Stt
v0.1.0语音转文字 - 使用OpenAI Whisper将音频文件识别为文字
⭐ 0· 398·0 current·0 all-time
by魏然@qiaotucodes
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
high confidencePurpose & Capability
The README and SKILL.md describe automatic handling of incoming audio and sending transcriptions to Feishu, but the included code (transcribe.py) is a simple CLI that only loads Whisper and writes output locally. No Feishu integration, webhooks, or required credentials are declared — inconsistent claims vs. implementation.
Instruction Scope
SKILL.md recommends 'automatic processing' and explicitly says transcriptions are sent to Feishu. The runtime instructions do not document how the agent will receive audio, authenticate to Feishu, or where credentials would come from. The actual code only processes local files and does not reference external endpoints.
Install Mechanism
No install spec is provided (instruction-only plus a small Python script). Dependencies are typical (openai-whisper, PyTorch, ffmpeg). Be aware model downloads (1–10GB) happen at first run and require network and disk space.
Credentials
The skill declares no required environment variables or credentials, yet SKILL.md describes sending results to Feishu (which would require tokens). The absence of declared env vars for any external service is inconsistent with the described automatic-sending behavior.
Persistence & Privilege
The skill does not request always:true and does not attempt to modify other skills or system settings. It runs locally as a CLI script; autonomous invocation is allowed by default but not unusual here.
What to consider before installing
This package is a simple, local Whisper-based transcription script — safe in the sense it only transcribes local files — but the documentation claims automatic processing and sending results to Feishu even though the code doesn't implement that. Before installing or enabling automated runs: 1) Confirm whether you want automatic network posting; if so, require the author to implement it and declare what env vars/tokens will be needed. 2) Expect large model downloads (1–10GB) and significant CPU/RAM or GPU needs; ensure sufficient disk space and bandwidth. 3) If running in a hosted agent, limit network/file permissions (run in isolated environment) until the Feishu/integration behavior is clarified. 4) Note the README/author metadata includes an obfuscated-looking identity string in the script — not proof of malice but worth verifying the source if you require provenance.Like a lobster shell, security has layers — review code before you run it.
latestvk97fn6b4dg4ndv92gnv7cp50dn81xzc2
License
MIT-0
Free to use, modify, and redistribute. No attribution required.