llm-trading-agent-security

What problem does it solve?

Autonomous trading agents can convert a single prompt injection, mis-signed transaction, or exposed key into irreversible financial loss. This Skill codifies layered security patterns and operational controls to reduce the risk of unauthorized transactions, slippage, and key compromise.

Core Features & Use Cases

• Prompt hygiene & input sanitization: Detect and reject injection patterns in external data before it reaches execution-capable prompts.
• Spend policy enforcement: Hard single-transaction and daily spend limits with recorded accounting to stop runaway transfers.
• Pre-send simulation & execution guards: Require simulation results and min_amount_out checks to prevent slippage and failing trades.
• Circuit breakers & audit logging: Halt on drawdowns, consecutive losses, or invalid state and ensure every decision is auditable.
• Wallet isolation & secret management: Use ephemeral hot wallets and secret managers; never embed primary treasury keys in agent code.
• Use Case: Audit and harden a DeFi arbitrage bot so it cannot send funds above set limits, must simulate trades successfully, and can be stopped automatically on abnormal losses.

Quick Start

Ask the agent to run a transaction-security audit and apply prompt sanitization, hard spend limits, simulation checks, circuit breakers, and secret-managed wallet isolation before any on-chain send.