The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions provide templates for the agent to execute shell commands using tools like ffmpeg, yt-dlp, and curl. These commands incorporate user-supplied inputs such as video URLs and timestamps. The provided templates use quoting for variables to reduce the risk of shell injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes external media and metadata from platforms like YouTube. This represents a potential surface for indirect prompt injection where instructions could be hidden in video titles or metadata.
Ingestion points: Metadata and video content fetched from external URLs via yt-dlp.
Boundary markers: None explicitly defined in the provided instruction set.
Capability inventory: Local file system writes to /tmp, network access via curl, and shell command execution.
Sanitization: No specific sanitization logic is described; the skill relies on the agent's internal safety guardrails for processing external tool output.
[EXTERNAL_DOWNLOADS]: The skill downloads content from well-known video hosting services (YouTube and Bilibili) and communicates with the vendor's official domain (verging.ai) to perform enhancement jobs. These activities are required for the skill's primary function.

video-enhancement