mirofish

MiroFish Simulation Skill

Leverage the MiroFish multi-agent prediction framework for data-driven strategic foresight.

When to Invoke

• User asks "what if" or "what would happen if" questions
• Scenario modeling or forecasting is needed
• Strategic decisions require simulation of outcomes
• Data-driven prediction beyond simple extrapolation

Overview

MiroFish uses a multi-agent architecture where specialized agents collaborate on prediction tasks:

• Data agents gather and preprocess input data
• Model agents run parallel simulations with different assumptions
• Synthesis agents aggregate results and identify consensus/divergence

Methodology

1. Define the question — What specific outcome are we trying to predict or model?
2. Identify variables — What inputs, assumptions, and constraints define the scenario?
3. Configure simulation — Set up agent roles, data sources, and model parameters
4. Run simulations — Execute parallel scenarios with varying assumptions
5. Synthesize results — Aggregate findings, identify patterns, flag uncertainties
6. Report — Deliver structured prediction report with confidence intervals

Output Format

## Prediction Report: [Question]

### Scenarios Modeled
- Scenario A: [description] — [probability] — [outcome]
- Scenario B: [description] — [probability] — [outcome]

### Key Variables
| Variable | Base Case | Optimistic | Pessimistic |
|----------|-----------|------------|-------------|

### Recommendation
[Action to take based on simulation results]

### Confidence & Caveats
[What we're confident about, what we're not, and why]

Setup

MiroFish can run as:

• Inline mode — Lightweight simulation within the current agent context
• Service mode — Full Docker-based service for heavy simulations

For service mode setup, see: https://github.com/666ghj/MiroFish