linkfox-temu-promotion-eu
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves promotion and product information from the Temu API and stores the responses in local JSON files for the agent to process. This creates a surface for indirect prompt injection where data from the external platform could influence the agent's instructions. 1. Ingestion points: External data enters the agent context via API responses handled by scripts like scripts/temu_proxy.py and eu_promotion_*.py endpoint scripts. 2. Boundary markers: The skill instructions recommend using jq to filter data, which helps scope the context, but the scripts do not implement strict boundary delimiters or instructions to ignore embedded commands. 3. Capability inventory: The skill has the capability to execute Python scripts, perform network operations to LinkFox and Temu domains, and read/write to the local filesystem. 4. Sanitization: While the skill parses structured JSON, it does not sanitize the text content within those structures before the agent reads them.
- [COMMAND_EXECUTION]: The skill executes multiple Python scripts to perform its primary functions, including API communication and local configuration management, which is typical for this type of integration.
- [EXTERNAL_DOWNLOADS]: The scripts/onboarding.py utility provides instructions for installing requests, qrcode, and pillow via pip to support user registration and payment workflows. These are well-known libraries and their usage is consistent with the skill's onboarding features.
- [DATA_EXFILTRATION]: The skill transmits commerce data and shop identifiers to LinkFox domains (linkfox.com) and Temu endpoints (temu.com). This network activity is necessary for the skill to perform its stated functions and targets vendor-owned or target-service infrastructure.
Audit Metadata