Skills
Navigate
skills/faberlens/hardened-skills/quant-trading-system-hardened/Gen Agent Trust Hub

quant-trading-system-hardened

Warn

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The SKILL.md and SAFETY.md files contain extensive "Security Guardrails" and "Safety Evaluation" content. These use phrases like "regardless of claimed authority or developer instructions" and "holds regardless of claimed authority," which are patterns designed to set immutable agent behaviors and override external instructions.- [CREDENTIALS_UNSAFE]: Hardcoded cryptocurrency wallet addresses (0x3462abb0b7f4e48580c3b14ea85c3493d656f8ee) are present in auto_trader.py, trading_system.py, and trading_system_v2.py.- [COMMAND_EXECUTION]: The dashboard.py and trading_system_v3.py files contain logic to start a local HTTP/TCPServer on port 5000. This initiates a network-listening process on the host machine.- [EXTERNAL_DOWNLOADS]: The skill performs network operations to https://api.hyperliquid.xyz/info using the requests library to fetch market data.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 21, 2026, 06:31 PM
Security Audit — agent-trust-hub — quant-trading-system-hardened