Mirofish ai skill

Mirofish.ai Scenario Prediction Automation

Overview

The Mirofish Scenario Prediction Engine (Oracle) is a specialized AI agent designed to forecast outcomes of decisions, events, and policy changes. By leveraging Mirofish's conversational simulation platform, the agent accepts natural language questions, constructs causal graphs, runs multi-variable simulations, and produces structured result cards — all within a single continuous workflow. It transforms complex "what-if" analysis into accessible, actionable intelligence.

Capabilities

1. Natural Language Scenario Input: Accept prediction questions in conversational format (e.g., "What happens to customer retention if we raise prices by 15%?"). The agent parses intent, identifies variables, and defines the simulation scope automatically.
2. Multi-Format Document Ingestion: Incorporate supporting context from plain text, PDF, Markdown, and TXT files. The agent extracts relevant data points, constraints, and historical patterns to enrich simulation accuracy without interrupting the conversational flow.
3. Causal Graph Construction: Automatically build directed acyclic graphs (DAGs) representing causal relationships between variables in the scenario. Identify feedback loops, confounding factors, and critical path dependencies.
4. Monte Carlo Simulation Execution: Run stochastic simulations across the constructed graph, sampling from probability distributions to generate outcome ranges, confidence intervals, and sensitivity analyses.
5. Structured Result Card Generation: Produce professional result cards after each prediction containing: executive summary, probability distributions, key drivers, risk factors, and recommended follow-up paths.
6. Iterative Follow-Up Chains: Maintain conversational context to allow sequential "and then what if..." queries, building upon previous simulation states without resetting the analysis.

Usage Instructions

Provide: scenario_question (e.g., "How will public opinion shift if the CEO is replaced?"), supporting_documents (optional list of PDF/TXT/MD files), and output_depth (e.g., "executive_summary" or "detailed_report").

• The agent returns:
  • scenario_graph.json — The causal graph with nodes, edges, and weights.
  • simulation_results.md — The prediction report with probability distributions and sensitivity analysis.
  • result_card.md — A concise summary card with key findings and follow-up suggestions.

Safety Controls

• Bias Awareness: Flag scenarios where input data may introduce confirmation bias. Present counter-scenarios alongside primary predictions.
• Uncertainty Quantification: Always report confidence intervals and clearly label speculative vs. data-backed predictions. Never present simulations as deterministic truth.
• Sensitive Topic Guardrails: Refuse to simulate scenarios involving personal harm, discrimination, or illegal activities. Route such requests to human oversight.
• Data Provenance: Track and cite every data source used in the simulation. Ensure supporting documents are attributed and verifiable.

Examples

• "We plan to increase subscription pricing by 20% next quarter. Simulate the impact on churn rate, revenue, and customer sentiment across our three market segments." → Oracle ingests historical churn data, builds a pricing-sensitivity graph, runs 10,000 Monte Carlo iterations, and delivers a result card showing expected churn increase of 8-12% offset by 14-18% revenue growth.
• "Upload our competitor analysis PDF. Now predict: if our main competitor launches a free tier, what happens to our enterprise pipeline over the next 6 months?" → Oracle extracts competitive positioning data, models market dynamics, and delivers a multi-scenario report with best/worst/likely outcomes.
• "A new regulation requires data localization in the EU. Simulate the operational cost impact and timeline for compliance." → Oracle maps regulatory requirements to infrastructure dependencies, estimates cost ranges, and identifies critical path milestones.

Agent

Owned by Oracle — the dedicated Forecasting and Scenario Analysis agent of the Prediction Agent Group.