The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to download Node.js from nodejs.org. This process involves verifying the binary's SHA256 checksum before execution, which is a security best practice for software installation.\n- [EXTERNAL_DOWNLOADS]: The skill installs the @iqinghu/qhkit package and sharp-cli from the official npm registry. It also provides a fallback to the Alibaba npm mirror for reliable access. These are standard and trusted sources for developer tools.\n- [DYNAMIC_EXECUTION]: The skill contains logic to resize images locally using Python (Pillow) or Node.js (sharp-cli) when files exceed the service's 10MB limit. These operations are limited to image manipulation and do not execute untrusted logic.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external media content.\n
Ingestion points: User-provided images, videos, and external URLs (e.g., Douyin) processed by video-inspire.\n
Boundary markers: The skill instructs the agent to confirm all generation tasks with the user before submission.\n
Capability inventory: Execution of qhkit CLI for media generation, local Python/Node.js scripts for image resizing.\n
Sanitization: Local scripts only perform image transformations; the CLI handles secure uploads to the vendor's service.

qinghu-ai