linkfox-temu-ads-us
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEPERSISTENCEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONOBFUSCATIONEXTERNAL_DOWNLOADS
Full Analysis
- [PERSISTENCE]: The skill instructions in
references/onboarding.mdadvise users to append environment variable exports to shell profiles such as~/.zshrcand~/.bashrcto persist authentication keys. - [DATA_EXFILTRATION]: The skill stores sensitive Temu access tokens in a local configuration file (
~/.linkfox/temu-access-tokens.json) and transmits them to vendor-controlled endpoints (tool-gateway.linkfox.com) for API proxying during normal operations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external API responses and saves them to local storage, creating a potential surface for indirect prompt injection if the stored data is processed by the agent without further validation.
- Ingestion points: API responses processed in
call_temu_apiandemit_result(scripts/_temu_common.py). - Boundary markers: None identified in instructions or scripts.
- Capability inventory: Capabilities include network POST requests, file system writes, and credential management.
- Sanitization: Standard JSON parsing is employed; no specialized prompt-level sanitization or escaping for ingested content was found.
- [OBFUSCATION]:
scripts/onboarding.pyutilizes a Base64-encoded string (_LOGIN_FIXED_UID) that decodes to a JSON object used in authentication headers for vendor services. - [EXTERNAL_DOWNLOADS]: The skill onboarding process recommends the manual installation of standard third-party Python packages including
requests,qrcode, andpillowto support login and QR code rendering functionalities.
Audit Metadata