human approval trading agent

The problem

Delegating trading to AI often leads to catastrophic "hallucination trades" or excessive risk-taking without oversight. Purely manual trading is too slow to catch opportunities, while fully autonomous bots lack human judgment during volatile market conditions.

What it does

• Generates structured trade proposals including entry, stop-loss, take-profit, and R-multiple calculations.
• Validates every proposal against pre-defined risk gates like max position size and total portfolio leverage.
• Enforces a mandatory "human-in-the-loop" protocol that prevents execution until an explicit APPROVE command is received.
• Monitors open positions and triggers new approval requests for any adjustments to stops or exits.
• Logs all gate checks, rationale, and timestamps for complete auditability of the trading process.

Why this beats prompting it yourself

Standard LLM prompts often result in the AI "assuming" consent or skipping risk calculations when under pressure. This skill builds in hard logical gates and specific response patterns that prevent the AI from ever placing an order autonomously, ensuring your capital is protected by both code and conscience.

Use cases

• Researching and staging crypto or equity trades while away from the terminal.
• Managing risk by forcing the AI to justify every trade against portfolio limits.
• Executing complex orders through a structured interface without manual data entry errors.

Known limitations

Requires an external brokerage API for final execution. Does not support high-frequency trading where approval latency would invalidate the entry price.