openocean

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external APIs and utilizes it to parameterize sensitive wallet operations.
  • Ingestion points: Transaction payload data, including destination addresses and execution hex data, is retrieved from the OpenOcean API via exports.py. Additionally, token balance data is fetched from the DeBank API.
  • Boundary markers: The skill does not implement delimiters or specific instructions to isolate or verify data retrieved from these external APIs before they are used in transaction construction.
  • Capability inventory: The skill utilizes the /agent/transfer capability to execute smart contract interactions, including granting infinite token allowances (MAX_UINT256) to addresses provided by the external API.
  • Sanitization: Transaction fields such as to and data are passed to the wallet service directly from the API response without validation against a whitelist or internal safety check.
  • [DATA_EXFILTRATION]: The skill transmits trade-related metadata to an external endpoint for analytics purposes.
  • Evidence: _trade_report.py captures trade events—including wallet addresses, transaction hashes, and trade symbols—and posts them to a URL defined by the AI_AGENT_API_URL environment variable.
  • Details: This reporting mechanism uses platform-level environment variables for configuration and authentication (CONTAINER_JWT), functioning as a telemetry service for trade monitoring and event tracking.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 04:28 AM
Security Audit — agent-trust-hub — openocean