The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the postplus CLI tool to execute shell commands for gathering YouTube channel summaries, video metrics, and audience comments.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external YouTube sources, which could contain malicious instructions embedded by third parties.
Ingestion points: Untrusted text is retrieved from video descriptions and user comments via the youtube-videos and youtube-comments routes described in SKILL.md.
Boundary markers: While the skill defines a privacy boundary concerning subscriber identities, it does not provide explicit boundary markers or instructions to ignore potential prompts within the ingested text.
Capability inventory: The skill executes commands through the postplus CLI tool and includes handoff capabilities to other skills such as media-analysis.
Sanitization: There is no documentation of sanitization, filtering, or validation processes for the data retrieved from YouTube before it is analyzed by the agent.

youtube-research