influencer-discovery
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a local helper script located at
scripts/connectors/youtube.pyto retrieve channel metrics. It executes this script via a shell command that takes a creator's handle as an argument to fetch subscriber counts and video statistics.\n- [PROMPT_INJECTION]: The skill is designed to process external information from social media profiles (such as bios and posts) which represents an indirect prompt injection surface. This data is incorporated into influencer profiles and reports.\n - Ingestion points: Influencer bios, about pages, and post content fetched from social platforms like YouTube, TikTok, and X (Twitter) via oEmbed URLs or local scripts.\n
- Boundary markers: The instructions guide the agent to use specific templates for organizing data but do not include explicit separators or instructions to ignore commands within the retrieved metadata.\n
- Capability inventory: The skill records results to local memory files and uses the
youtube.pyscript for data retrieval.\n - Sanitization: The skill relies on public metadata and provides structured templates for the agent to use when compiling reports.
Audit Metadata