worldthreatmodelharness_skill

Overview

This skill is a persistent world-model system that maintains 11 horizon-specific analyses from 6 months to 50 years for adversarial testing of ideas, strategies, and investments. It runs parallelized agents and specialized modules (RedTeam, FirstPrinciples, Council, Research) to stress-test inputs against every horizon. The system supports quick checks, standard analysis, and deep research tiers depending on decision criticality.

How this skill works

The harness stores a deep (~10-page) model for each time horizon and routes user requests to the appropriate workflow: testing, updating, or viewing models. When you request a test, it runs multilayered analyses across all horizons using synthesis agents plus adversarial and first-principles modules; tiers determine depth and runtime. Model updates trigger research agents that refresh horizon content and persist results to the WorldModels directory.

When to use it

• Stress-test an idea, strategy, or investment against multiple futures
• Assess long-term viability or resilience across time horizons
• Update or refresh world models with new research or intelligence
• Quickly view the current state and last-updated dates of all horizons
• Prepare for high-stakes decisions that require adversarial scenario analysis

Best practices

• Choose the tier based on risk: Fast for quick checks, Standard for most decisions, Deep for major investments
• Provide clear objectives and assumptions to improve FirstPrinciples decomposition
• Iterate: run tests, apply findings, then refresh models before re-testing
• Use the view workflow to confirm model currency and last-updated metadata before relying on results
• Store any customizations in the designated SKILLCUSTOMIZATIONS path so runs respect local preferences

Example use cases

• Validate a corporate strategy for resilience across 6-month to 50-year horizons before rollout
• Stress-test an investment thesis for geopolitical and technological tail risks
• Run adversarial analysis on a policy proposal to uncover blind spots and failure modes
• Refresh the 10-year and 20-year models with new research when major domain shifts occur
• Perform a quick Fast-tier sanity check before scheduling a Deep-tier review for board decisions