linkfox-chuhaijiang-tiktok-product

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCECOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted product descriptions, store names, and user reviews from TikTok Shop via external API responses.
  • Ingestion points: API responses processed in scripts/chuhaijiang_product_search.py, scripts/chuhaijiang_product_detail.py, scripts/chuhaijiang_product_reviews.py, and scripts/chuhaijiang_product_image_search.py.
  • Boundary markers: Absent; the scripts do not wrap ingested data in delimiters or include specific instructions to ignore embedded prompts.
  • Capability inventory: The scripts possess capabilities for local file system writes (writing responses to the linkfox/ directory) and network operations to vendor APIs.
  • Sanitization: Absent; content from the API is saved and displayed without filtering or escaping.
  • [PERSISTENCE]: The skill's onboarding documentation (references/onboarding.md) provides manual instructions for users to persist API keys by modifying shell configuration files such as ~/.bashrc and ~/.zshrc.
  • [COMMAND_EXECUTION]: The skill implements its functionality through Python scripts that perform network requests to vendor-owned domains (linkfox.com) and write market data to the local file system in the execution directory.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 06:50 AM