The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install the Node.js runtime from nodejs.org if it is missing from the environment. This is a well-known and trusted service. The process includes a security step to verify the downloaded file's integrity using SHA256 checksums before extraction, which is a defensive best practice.
[COMMAND_EXECUTION]: The skill executes the qhkit CLI tool (installed via npm as @iqinghu/qhkit) to perform its core functions, including video generation, task status monitoring, and credit estimation. This is the intended and legitimate behavior of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection because it ingests untrusted user data to generate content.
Ingestion points: The prompt and uploadedImages parameters in the qhkit video generate and qhkit video-quick generate commands within SKILL.md.
Boundary markers: Untrusted inputs are encapsulated within JSON strings passed as arguments to the CLI, providing structural separation from the shell command itself.
Capability inventory: The skill can execute the qhkit binary (subprocess), read local files (for image uploads), and perform network operations (communicating with the video generation API).
Sanitization: No explicit sanitization or filtering of the user-provided prompt or image file paths is performed within the instruction logic.

linkpix-ecom-video