The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads a Node.js binary and its corresponding checksum file from the official nodejs.org website. It uses a verification pipeline involving curl, grep, and sha256sum to validate the file integrity before extracting and executing the contents. It also suggests using npx for dynamic execution if global installation fails.
[EXTERNAL_DOWNLOADS]: Fetches the @iqinghu/qhkit package and its dependencies from the official NPM registry or a well-known mirror (npmmirror.com) maintained by Alibaba.
[COMMAND_EXECUTION]: Executes system-level shell commands to manage directories, install packages via npm, configure the environment, and run the qhkit CLI tool for image generation tasks.
[PERSISTENCE]: The skill instructs the agent to modify the user's shell profile (~/.bashrc) by appending environment variable updates. This ensures that the manually installed Node.js environment is persistently available in future terminal sessions, which is a common persistence technique.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a surface where untrusted user-supplied data is interpolated into command execution logic.
Ingestion points: Processes user-provided text descriptions (prompts) and file paths for product images within the qhkit image generate command in SKILL.md.
Boundary markers: Parameters are structured within JSON objects, providing basic syntactic separation; however, the skill lacks explicit instructions for the agent to ignore or delimit potentially malicious instructions embedded in the user prompt.
Capability inventory: The skill enables the agent to execute shell commands, perform file system writes during setup, and initiate network requests through the tool's integrated API.
Sanitization: No explicit sanitization, validation, or escaping of user-provided prompt content or image metadata is mentioned before the data is passed to the generation tool.

linkpix-scene-image