The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads Node.js binaries from nodejs.org and installs the @iqinghu/qhkit CLI tool from the NPM registry. It also uses the Aliyun npmmirror as a fallback for connectivity issues.
[COMMAND_EXECUTION]: The skill executes shell commands to perform AI video cloning tasks, manage configuration, and set up the execution environment.
[DYNAMIC_EXECUTION]: The instructions include logic to generate and execute short Python and Node.js snippets (e.g., using python -c and npx sharp-cli) to compress user-provided images that exceed size limits.
[PERSISTENCE]: The installation instructions suggest appending path configuration to the user's shell profile (~/.bashrc) to ensure the Node.js binaries are available in future sessions.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external video URLs (TikTok, Douyin, etc.), which creates an attack surface where adversarial instructions embedded in video metadata or scripts could influence agent behavior.
Ingestion points: The resourceUrl parameter in SKILL.md captures external video links.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded in the external content.
Capability inventory: The skill can execute shell commands (qhkit), install packages via npm, and run dynamic code via python or npx.
Sanitization: No explicit sanitization or filtering is performed on the data retrieved from external links.

linkpix-seedance-2-clone