The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from YouTube, including video comments and captions, which could contain malicious instructions designed to influence agent behavior.
Ingestion points: Untrusted data enters the agent context through tools like youtube_web_v2_video_comments, youtube_web_v2_video_comment_replies, and youtube_web_v2_video_captions (referenced in SKILL.md and references/sandbase-api-map.md).
Boundary markers: The instructions do not include specific delimiters or warnings for the agent to ignore instructions embedded in the external content.
Capability inventory: The skill uses sandbase_call_tool to interact with the YouTube API and retrieve content (SKILL.md).
Sanitization: There is no mention of sanitizing or filtering the retrieved data before it is processed by the agent.
[NO_CODE]: The skill consists exclusively of markdown documentation and instructions for using existing platform tools. It does not include any scripts, binaries, or dependency configuration files.

youtube-research