reverse-engineer-niche

Pass

Audited by Gen Agent Trust Hub on Nov 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, including YouTube video titles, transcripts, and metadata, as part of the research workflow (documented in SKILL.md steps 9 and 10). While no explicit boundary markers or sanitization logic are defined to prevent the agent from processing instructions potentially embedded in this data, the skill's activities are focused on data analysis and report generation.
  • [DYNAMIC_EXECUTION]: The HTML template provided in the skill includes a small JavaScript block. This script is used for keyboard navigation (scrolling between pages) within the generated report and does not perform sensitive operations or network requests.
  • [EXTERNAL_DOWNLOADS]: The skill and its generated output reference resources from well-known services like Google Fonts and YouTube's image CDN. It also includes documentation suggesting the installation of further research tools from the vendor's GitHub repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 21, 2026, 07:33 PM