The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides commands to download Node.js binaries and checksum integrity files directly from the official nodejs.org domain to ensure the environment meets the execution requirements.
[REMOTE_CODE_EXECUTION]: The skill uses package managers such as npm and npx to download and execute code at runtime, specifically for the primary toolkit (@iqinghu/qhkit) and auxiliary image processing libraries (sharp-cli).
[COMMAND_EXECUTION]: The skill executes shell commands via the qhkit binary to interface with the LinkPix video generation service, including tasks for task submission, status polling, and model estimation.
[PERSISTENCE]: The documentation instructs users to modify their ~/.bashrc file to persistently include the local Node.js installation in the system PATH variable, ensuring the tools remain available across different shell sessions.
[DYNAMIC_EXECUTION]: The skill generates and executes temporary Python or Node.js code snippets at runtime to compress images larger than 10MB, using libraries like Pillow or sharp-cli for local processing.
[INDIRECT_PROMPT_INJECTION]: By processing untrusted user data (text prompts and image files) through generative AI models, the skill maintains a surface for indirect prompt injection attacks.
Ingestion points: User-provided video prompts, uploaded product images, and media reference URLs.
Boundary markers: Uses structured JSON for CLI arguments to separate operational instructions from user-supplied data strings.
Capability inventory: File system read/write access (for asset processing), network communication (for uploads and API calls), and subprocess execution.
Sanitization: There are no explicit instructions or filters to sanitize user-supplied creative prompts before they are passed to the video generation models.

linkpix-tiktok-viral-video