shengjiang-viral-breakdown
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as video transcripts, social media comments, and user-provided documents. It lacks explicit instructions to wrap this content in boundary markers or perform sanitization, creating a vulnerability surface where malicious instructions embedded in the source material could manipulate the agent. * Ingestion points: User-provided video URLs, local video files, PDFs, and scraped social media comments as described in SKILL.md. * Boundary markers: None identified in the skill's instructions. * Capability inventory: Subprocess execution via FFmpeg for frame extraction, file system operations for report generation, and coordination with the shengjiang-research companion skill. * Sanitization: No explicit validation or filtering of external data is specified.
- [COMMAND_EXECUTION]: The skill utilizes FFmpeg or similar video tools to extract keyframes from untrusted media files, which involves executing shell commands.
- [EXTERNAL_DOWNLOADS]: The skill relies on a companion skill, shengjiang-research, from the same vendor's repository to download video content and metadata from external platforms.
Audit Metadata