linkfox-temu-cancel-order-us
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE]: The skill automatically saves full API responses, which may include PII and sensitive order details, to a
linkfox/directory within the current working directory. This documented behavior creates a risk of accidental data exposure if the workspace is committed to version control or shared with third-party tools. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from the Temu API and generates summaries for the agent, creating a surface for indirect prompt injection.
- Ingestion points: API responses retrieved from
tool-gateway.linkfox.comvia thecall_temu_apifunction in_temu_common.py. - Boundary markers: None identified in the instructional content or scripts to separate data from agent instructions.
- Capability inventory: The skill possesses file writing capabilities (
emit_result), network access to vendor domains, and local script execution. - Sanitization: Data is parsed as JSON but not specifically sanitized or filtered for instruction-injection patterns before being presented to the agent.
- [EXTERNAL_DOWNLOADS]: The skill performs network operations targeting
linkfox.comand its subdomains for gateway proxying, feedback, and account management. As these resources are owned by the skill vendor (linkfox-ai), they are documented as necessary functional components. - [COMMAND_EXECUTION]: The skill relies on executing multiple local Python scripts to interact with the LinkFox gateway and manage local token storage. This is a primary architectural feature restricted to the skill's distributed script files.
Audit Metadata