The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using npm exec to interact with the KickArt CLI (kickart-open-cli) for task submission and status queries.\n- [EXTERNAL_DOWNLOADS]: Fetches the vendor-owned package @volcengine/kickart-open-mcp from the official npm registry (https://registry.npmjs.org) at runtime.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input, including descriptive prompts and external media URLs, which are then passed to the command-line interface.\n
Ingestion points: Media URLs and local paths collected in references/视频参考指南.md and references/替换图片指南.md; natural language descriptions collected in references/视频元素替换指南.md.\n
Boundary markers: CLI arguments are wrapped in double quotes; however, there are no specific filters to prevent instructions within the prompt string from influencing the agent's logic if processed further.\n
Capability inventory: Execution of shell commands via npm exec for task management (SKILL.md).\n
Sanitization: The instructions explicitly advise against manual string concatenation for shell commands to mitigate injection risks.

byted-kickart-video-elements-replace