Video Captions

Generate professional captions and subtitles with multi-engine transcription, word-level timing, styling presets, and burn-in.

This skill appears coherent for captioning tasks, but review these before using: 1) It expects ffmpeg and a Whisper CLI on PATH — ensure you trust those binaries and their versions. 2) The docs recommend pip installs (openai-whisper, whisper-timestamped, stable-ts, etc.); install packages from trusted sources and consider a virtualenv or container. 3) Local processing is the default (best for privacy). Only supply cloud API keys (AssemblyAI / Deepgram) if you consent to sending video/audio to those services. 4) Example ffmpeg burn-in commands depend on fonts and may require installing or mapping fonts on your system. 5) If you plan to let an autonomous agent run this skill, be aware it could process any video files the agent is given — limit scope and run in an isolated environment if you have sensitive content.

Type: OpenClaw Skill Name: video-captions Version: 1.0.1 The skill bundle is designed for legitimate video captioning tasks, utilizing tools like `ffmpeg` and `whisper`. It explicitly states that processing is local by default and cloud APIs are opt-in, requiring user-provided keys, which mitigates concerns about hidden data exfiltration. However, the extensive use of `ffmpeg` and `whisper` commands across `SKILL.md`, `engines.md`, `formats.md`, `platforms.md`, and `styling.md` introduces a significant shell injection vulnerability risk. If the AI agent constructs these commands using unsanitized user input (e.g., for filenames or style parameters), a malicious user could execute arbitrary commands on the host system. While there is no evidence of intentional malicious behavior within the skill's instructions or code, this critical vulnerability potential warrants a 'suspicious' classification.