The Agent Skills Directory

[PERSISTENCE]: The instructions in SKILL.md direct the agent to modify the user's shell profile (~/.bashrc) by appending the installation path of the Node.js runtime. This mechanism ensures that environment changes persist across subsequent terminal sessions.
[EXTERNAL_DOWNLOADS]: The skill provides procedures to download the Node.js binary from the official nodejs.org domain and the @iqinghu/qhkit package from the npm registry or its Alibaba mirror (registry.npmmirror.com). These sources are recognized as trusted or well-known services.
[REMOTE_CODE_EXECUTION]: The skill suggests a fallback mechanism using npx to execute the @iqinghu/qhkit package directly from the npm registry, which involves the dynamic download and execution of remote code at runtime.
[COMMAND_EXECUTION]: The skill frequently invokes the qhkit command-line tool with various arguments to generate scripts, create images, and check task status.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied data such as product names and image paths, which are subsequently interpolated into shell commands. This creates a potential surface for indirect prompt injection if the inputs contain malicious control sequences.
Ingestion points: The productName, pointDescription, and uploadedImages fields within the JSON payloads in SKILL.md.
Capability inventory: Extensive use of shell command execution for installation (npm, curl, tar) and operation (qhkit).
Boundary markers: Parameters are structured as JSON strings before being passed to the CLI tools, providing basic architectural separation.
Sanitization: The instructions do not specify any validation or escaping routines for user-supplied text before it is processed by the external tools.

qinghu-storyboard