The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external media (source videos, reference images) and user-supplied prompts. While this represents a theoretical surface for indirect prompt injection via media metadata or visual content, the skill incorporates specific safeguards, such as explicit preservation lists and manual review gates (video QA), to mitigate accidental compliance with embedded instructions.
[DATA_EXFILTRATION]: The skill uses media upload and processing capabilities. It includes a specific instruction to avoid placing private media in public prompts or receipts, demonstrating a security-conscious design for handling user assets.
[COMMAND_EXECUTION]: No shell commands, dynamic context injection (!command syntax), or executable scripts are present in the skill definition.
[OBFUSCATION]: The content is written in clear, standard Markdown with no evidence of encoded strings, hidden characters, or homoglyph attacks.

Video Editing