youtube
Pass
Audited by Gen Agent Trust Hub on Jul 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill performs specialized tasks by executing local Python scripts within a managed sandbox environment. This architecture allows for complex logic, such as thumbnail styling and SEO packaging, while maintaining isolation from the host system.
- [EXTERNAL_DOWNLOADS]: The skill fetches external data from YouTube, specifically transcripts and video metadata, through the
youtube_toolkit. This is an essential function for providing summaries and SEO assets and is performed through sanctioned platform integrations. - [DATA_EXFILTRATION]: While the skill processes user-uploaded images and video identifiers, there is no evidence of unauthorized data transmission to external domains. The scripts explicitly use internal file identifiers (
file_id) to manage assets within the conversation context. - [PROMPT_INJECTION]: The skill utilizes internal prompts for tasks like generating thumbnail concepts and SEO copy. These prompts are narrowly focused on the specific output formats (JSON) and do not include instructions that would encourage the agent to bypass its standard safety protocols or ignore user constraints.
Audit Metadata