1688-product-search

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The smart_recommend.py script invokes other Python scripts within the skill using subprocess.run() to modularize search and recommendation logic.
  • [CREDENTIALS_UNSAFE]: The skill implements an authentication caching mechanism that stores OAuth2 access and refresh tokens in a JSON file at a predictable local directory (~/.openclaw/workspace/skills/.1688_token_cache.json) to share authentication state between related vendor skills.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by downloading images from user-provided URLs for the 'image-search' feature. Ingestion points: Data enters the environment via URL-based image downloads in product_search.py and image_search_handler.py. Boundary markers: None identified; data is processed and sent to the 1688 API as structured payloads. Capability inventory: Includes network operations, file system access for caching and temporary storage, and subprocess execution. Sanitization: External images are processed, converted to RGB, and compressed using the Pillow library before being re-uploaded to the platform.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 04:27 PM
Security Audit — agent-trust-hub — 1688-product-search