The Agent Skills Directory

[COMMAND_EXECUTION]: The script douyin.js spawns subprocesses for ffmpeg and ffprobe to handle media conversion and metadata extraction. While it uses structured argument arrays with child_process.spawn (which is generally safer than exec), the capability to execute external binaries on the system is a significant permission.
[EXTERNAL_DOWNLOADS]: The skill performs multiple network requests to external domains, including Douyin (douyin.com) for content fetching and SiliconFlow (api.siliconflow.cn) for audio transcription. These operations are necessary for the skill's purpose but involve sending and receiving data from external services.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data retrieved from external URLs, which represents a potential attack surface for indirect prompt injection if the processed content is later interpreted by the agent.
Ingestion points: External video metadata and page content parsed from Douyin URLs in douyin.js via the parseShareUrl function.
Boundary markers: None identified in the script logic to distinguish between data and instructions.
Capability inventory: The skill has the capability to write to the file system (fs.writeFileSync), execute subprocesses (ffmpeg), and perform network requests (https.request).
Sanitization: The script performs basic sanitization on video titles (e.g., removing path-unsafe characters) before using them in filenames, but it does not specifically sanitize content against prompt injection patterns.

douyin-downloader