Clawock: Make Your AI Agents Debate Trades Before Code Settles

Where this fits

You're running a Claude Code agent that manages a small brokerage account. The agent suggests a buy order for a tech stock based on a news spike, but you're worried it didn't weigh the downside. Clawock solves this by making every trade proposal go through a debate phase—the agent must argue for and against—and then a separate, deterministic code module executes the trade. This keeps the model's creativity away from the final action, reducing catastrophic errors.

Why agents benefit

• Forced deliberation: The agent must generate both a bullish and bearish thesis before any trade is considered.
• Deterministic settlement: The actual trade is executed by code that uses rules and can't drift or hallucinate.
• Runtime agnostic: The skill works with OpenClaw, Claude Code, Codex, DeepSeek Harness, or your own agent harness.
• Transparency: Every decision and its outcome is published, so you can audit performance.
• Error containment: Debate surfaces edge cases (slippage, liquidity) that a single-pass agent might miss.

Practical scenarios

Solo Developer Running a Personal Trading Bot

You have an agent that watches markets and proposes trades. Instead of letting it fire off orders, wrap it with Clawock. The agent debates each candidate trade against opposing factors, and only then does code execute—keeping your account safe from model overconfidence.

Multi-Agent Research Team for a Quant Fund

Use Clawock as the shared decision layer across multiple agents with different strategies. Each agent argues its position, and the code settles with rules you define (e.g., max position size, stop-loss). The result is a disciplined execution pipeline with a full debate log.

Backtesting with Simulated Trades

Test your agent's decision process on historical data. Clawock logs the debate and the settled action, letting you compare against actual market outcomes. This gives you a feedback loop to tune your agent's reasoning without risking real capital.

Add it to your agent workflow

Install with a single command:

npx skills add KCNyu/clawock

After adding, you can integrate the workflow into your agent. For example, in Claude Code, you'd load the skill and instruct your agent to call the debate module before any trade action:

from clawock import debate_and_settle

# Your agent's trade proposal
proposal = {"symbol": "TSLA", "side": "buy", "size": 10}

# Clawock forces debate and returns a settled order
order = debate_and_settle(proposal, ruleset="conservative")
print(order)

The skill handles the debate prompts, the settlement rules, and the execution hook.

Compare before adopting

Before integrating, assess:

• Quality signals: The repo has 10 stars, a PyPI/npm package, and a test harness—check the commit history and recent activity.
• Maintenance freshness: Verify that the skill is updated for current agent runtimes and the debate logic reflects market feedback.
• Alternatives: Consider other finance-agent frameworks (e.g., FinRL, TradingAgents) and see if they offer the same code-settles enforcement.
• Workflow fit: Does it match your agent's existing decision loop? Clawock is opinionated—you need to wrap your trade functions around it.

Why it is worth tracking

Clawock stands out because it forces a debate and settles in code, a pattern that directly addresses the 'hallucination on execution' problem in agentic finance. With packages on both PyPI and npm, it's designed for multi-runtime adoption. The repository publishes results, win or loss, which is rare and builds trust. Keep an eye on its star growth and community forks—if the debate logic evolves to cover more asset classes, it becomes a must-evaluate skill for any developer putting agents near real money. For a deeper look at similar skills, check the OpenAgentSkill entry at https://www.openagentskill.com/skills/kcnyu-clawock.