The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a central CLI utility bin/wb to orchestrate various tasks including video rendering, audio mixing, and image processing. This involves executing system commands such as ffmpeg, ffprobe, node, and python to process local media assets.
[EXTERNAL_DOWNLOADS]: The skill interacts with several external APIs to fulfill its primary functions: fetches official SVG logos from Wikimedia Commons, utilizes ByteDance (Volcengine) and Xiaomi (MiMo) APIs for high-quality text-to-speech synthesis, and leverages the Codex CLI for generating image stickers. These services are well-known technology vendors.
[DATA_EXPOSURE]: The config.json file contains a hardcoded local file path in the tts.edge.python field that references a specific user directory (C:/Users/wwwzh/). While this exposes a local username, it is a minor information leak typical of local configuration files and does not pose a direct threat to the execution environment.
[DYNAMIC_EXECUTION]: The skill's workflow involves generating and executing Node.js scripts (scenes.js) which define the video content and layout. This is the intended design of the tool to allow for programmatic video creation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided topics to generate video scripts and image prompts. While this presents an injection surface, the skill provides extensive instructions and reference guidelines to ensure the agent maintains context and adheres to safety and quality standards. The risk is considered low and managed by standard model guardrails.

whiteboard-video-factory