ecommerce-reverse-product-image-search-api

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes subprocess.run within scripts/response_io.py and scripts/step_3_5_junglescout.py to invoke package-local Python scripts. While these operations are used for legitimate task chaining, they establish an execution pattern that involves the local shell.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted product information (such as titles and categories) from various marketplaces including Amazon, Walmart, and TikTok. (1) Ingestion points: Untrusted data enters the agent context through marketplace JSON files loaded in scripts/step_4_merge_rank.py and scripts/step_3_5_junglescout.py. (2) Boundary markers: No specific delimiters or instructions are used to separate external data from system prompts. (3) Capability inventory: The skill has access to network requests via urllib.request, file system writes, and subprocess execution. (4) Sanitization: The implementation lack explicit filtering or sanitization of ingested string values before processing.
  • [DATA_EXFILTRATION]: A file upload utility in scripts/nexscope_paths.py uses the oss2 library to transmit local files to an external Alibaba Cloud OSS bucket. This feature is intended for deliverable management within the vendor's ecosystem but represents a mechanism for transferring local data to external storage.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:14 AM
Security Audit — agent-trust-hub — ecommerce-reverse-product-image-search-api