polymarket-trading

Pass

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: SAFECREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill accesses a sensitive private wallet key stored at ~/.openclaw/blockrun/wallet.key or via the BLOCKRUN_WALLET_KEY environment variable. This key is used to sign financial transactions and pay for service fees.
  • [DATA_EXFILTRATION]: All trading traffic is routed through a remote egress server in Tokyo managed by the vendor. This entails sending transaction signatures and order details to an external network endpoint.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It ingests market data (e.g., token IDs, market conditions) from the blockrun_predexon_markets tool and subsequently performs high-consequence financial actions via blockrun_polymarket based on that data.
  • Ingestion points: Data retrieved from blockrun_predexon_markets as described in Step 4 of the flow.
  • Boundary markers: None present; the instructions do not specify delimiters or warnings to ignore instructions embedded in market data.
  • Capability inventory: The blockrun_polymarket tool can execute buys, sells, withdrawals, and fund transfers.
  • Sanitization: No evidence of input validation or sanitization for market-sourced data before it is passed to the trading tool.
  • [COMMAND_EXECUTION]: The instructions direct the agent to execute the blockrun_polymarket CLI tool for managing the entire trading lifecycle, including wallet setup and fund withdrawals.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 20, 2026, 10:16 AM
Security Audit — agent-trust-hub — polymarket-trading