The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses dynamic loading and runtime code generation to function. Specifically, scripts/qa.mjs and scripts/subtitles.mjs utilize the import() function to execute timeline.ts files that are created or modified during the workflow. Additionally, the workflow involves using sed to generate new JavaScript files (audio/episodes/<期>.mjs) which are subsequently executed by Node.js.\n- [COMMAND_EXECUTION]: The workflow relies on executing several external CLI tools via execFileSync, spawnSync, and shell commands. This includes ffmpeg for audio/video processing, npx remotion for rendering frames, and various file management commands like mv and sed.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided topics to drive the entire content creation process, which creates a potential vector for indirect prompt injection.\n
Ingestion points: The skill accepts a "topic" (选题) from the user or from a local Markdown file (materials/09-黑金光影风/选题.md) which is then used to generate scripts and research.\n
Boundary markers: There are no explicit delimiters or safety warnings mentioned in the instructions for the agent when interpolating the user topic into research or script-writing prompts.\n
Capability inventory: The skill has the ability to write files, execute shell commands (ffmpeg, npx, sed), and dynamically load/execute generated TypeScript and JavaScript files.\n
Sanitization: The provided scripts do not show any sanitization or validation of the user-provided topic before it is used to generate executable code and prompts.

weilanx-creator-blackgold-video