The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of the arkcli tool to perform generation tasks, query model capabilities, and manage asynchronous task states. This execution is central to the skill's purpose and follows a structured multi-step workflow.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input in the form of text prompts and media files, which are used to drive the generation process. 1. Ingestion points: User-provided content is ingested via the <prompt> positional argument and the --input parameter for referencing images, videos, and audio files, as detailed in SKILL.md and references/arkcli-gen.md. 2. Boundary markers: The instructions do not define specific delimiters or escaping mechanisms to isolate user-supplied strings from the command execution context, although the use of quotes is implied for prompts. 3. Capability inventory: The skill has the ability to execute shell commands, perform network operations via the CLI tool, and write files to the local directory (e.g., using the --save-to flag). 4. Sanitization: The skill does not perform local sanitization of inputs; instead, it relies on the backend service's content safety filters (such as ContentRiskBlocked) to mitigate malicious or sensitive content.
[EXTERNAL_DOWNLOADS]: The skill is configured to automatically download generated media from Volcengine's services to the local filesystem using the arkcli tool. These downloads are directed to the current working directory by default and are a core part of the content generation lifecycle.

arkcli-gen