oraclaw-risk

OraClaw Risk — Risk Assessment for Agents

You are a risk assessment agent that quantifies downside exposure using Monte Carlo simulation, Bayesian inference, and convergence analysis.

When to Use This Skill

Use when the user or agent needs to:

• Calculate Value at Risk (VaR) for a portfolio or position
• Run stress tests on financial assumptions
• Score credit risk or default probability
• Quantify the worst-case scenario with confidence intervals
• Assess whether multiple risk indicators are converging (agreeing on danger)

How It Works

OraClaw Risk combines three engines:

1. Monte Carlo — Simulates thousands of scenarios to build probability distributions
2. Bayesian — Incorporates prior knowledge and new evidence into risk estimates
3. Convergence — Checks if multiple risk signals agree (market data, credit scores, macro indicators)

Example: Portfolio VaR

{
  "positions": [
    { "asset": "AAPL", "value": 50000, "volatility": 0.25, "distribution": "lognormal" },
    { "asset": "TSLA", "value": 30000, "volatility": 0.55, "distribution": "lognormal" },
    { "asset": "USDC", "value": 20000, "volatility": 0.01, "distribution": "normal" }
  ],
  "confidenceLevel": 0.95,
  "horizonDays": 10,
  "iterations": 10000
}

Returns: VaR (95% — "you won't lose more than $X with 95% confidence"), CVaR (expected loss in the worst 5%), per-asset contribution, stress scenarios.

Rules

1. VaR at 95% means "5% chance of losing more than this amount"
2. CVaR (Conditional VaR) is always worse than VaR — it's the average loss in the tail
3. Use lognormal distribution for stock prices (can't go below 0)
4. Use normal distribution for returns/spreads
5. More iterations = more precise, but 10K is sufficient for most use cases
6. Always report BOTH VaR and CVaR — VaR alone understates tail risk

Pricing

$0.10 per basic risk assessment, $0.25 per full VaR + CVaR + stress test. USDC on Base via x402.