brand-protection-tiktok

Pass

Audited by Gen Agent Trust Hub on Oct 4, 2026

Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [METADATA_POISONING]: The skill claims to be for TikTok Shop brand protection, yet the functional code and templates are targeted at Amazon sellers. This mismatch could lead to the application of incorrect platform policies or URLs. \n
  • Evidence: scripts/detector.py utilizes Amazon-specific asin identifiers and references 'Amazon Brand Registry' in comments and user-facing strings.\n
  • Evidence: scripts/templates.py includes hardcoded URLs for https://brandregistry.amazon.com/ and references Amazon's anti-counterfeiting policies.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted product reviews, which could contain malicious instructions. \n
    1. Ingestion points: reviews dictionary ingested by detect_counterfeit_signals in scripts/detector.py.\n
    1. Boundary markers: Absent in the report generation logic.\n
    1. Capability inventory: No high-privilege tools are requested in the skill frontmatter, but the report output is intended for agent processing.\n
    1. Sanitization: Absent; content is analyzed for keywords without structural sanitization to prevent the injection of instructions.\n- [EXTERNAL_DOWNLOADS]: The skill is installed from a remote vendor repository. \n
  • Evidence: npx skills add nexscope-ai/eCommerce-Skills in SKILL.md.\n- [COMMAND_EXECUTION]: The skill relies on running local Python scripts to perform detection and template generation. \n
  • Evidence: python3 scripts/detector.py and python3 scripts/templates.py usage instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 4, 2026, 09:28 AM