ecommerce-temu-category-search

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The script scripts/temu_category_search.py transmits the NEXSCOPE_API_KEY via the Authorization header to a remote endpoint defined by the NEXSCOPE_PROXY_BASE environment variable. This is functional behavior for the vendor's proxy client but involves the network transmission of credentials.
  • [COMMAND_EXECUTION]: The script scripts/temu_category_search.py performs multiple file system operations, including creating nested directories and writing JSON response files to ./nexscope or ~/nexscope based on environment variables and current working directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes category names and descriptions retrieved from external Temu data synced to a local database. This content is interpolated into the agent's context without sanitization or protective boundary markers.
  • Ingestion points: API response data parsed and printed by scripts/temu_category_search.py.
  • Boundary markers: None identified in the script output or skill instructions to distinguish data from instructions.
  • Capability inventory: Local file writes (caching and session logging), directory creation, and network communication via the standard library.
  • Sanitization: None; the script prints the raw or summarized JSON response directly to standard output.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:14 AM
Security Audit — agent-trust-hub — ecommerce-temu-category-search