playbooks

home / skills / danielmiessler / personal_ai_infrastructure / worldthreatmodelharness

WorldThreatModelHarness skill

/Releases/v3.0/.claude/skills/WorldThreatModelHarness

This skill orchestrates eleven horizon world models to stress-test ideas, strategies, and investments against future scenarios using adversarial analysis.

npx playbooks add skill danielmiessler/personal_ai_infrastructure --skill worldthreatmodelharness

Review the files below or copy the command above to add this skill to your agents.

Files (6)
SKILL.md
3.6 KB
---
name: WorldThreatModelHarness
description: >
  Persistent world model system across 11 time horizons (6mo→50yr) for adversarial analysis of ideas,
  strategies, and investments. USE WHEN threat model, world model, test idea, test strategy, future analysis,
  test investment, how will this hold up, test against future, update world models, view world models,
  time horizon analysis, adversarial future test, stress test idea.
implements: Science
science_cycle_time: macro
context: fork
---

# World Threat Model Harness

A system of 11 persistent world models spanning 6 months to 50 years. Each model is a deep (~10 page)
analysis of geopolitics, technology, economics, society, environment, security, and wildcards for that
time horizon. Ideas, strategies, and investments are tested against ALL horizons simultaneously using
adversarial analysis (RedTeam, FirstPrinciples, Council).

## Workflow Routing

| Trigger | Workflow | Description |
|---------|----------|-------------|
| "test idea", "test strategy", "test investment", "how will this hold up", "stress test", "test against future" | `Workflows/TestIdea.md` | Test any input against all 11 world models |
| "update world model", "update models", "refresh models", "new analysis" | `Workflows/UpdateModels.md` | Refresh world model content with new research/analysis |
| "view world model", "show models", "current models", "model status" | `Workflows/ViewModels.md` | Read and summarize current world model state |

## Tier System

All workflows support three execution tiers:

| Tier | Target Time | Strategy | When to Use |
|------|-------------|----------|-------------|
| **Fast** | ~2 min | Single agent synthesizes across all models | Quick gut-check, casual exploration |
| **Standard** | ~10 min | 11 parallel agents + RedTeam + FirstPrinciples | Most use cases, good depth/speed balance |
| **Deep** | Up to 1 hr | 11 parallel agents + per-horizon Research + RedTeam + Council + FirstPrinciples | High-stakes decisions, major investments |

**Default tier:** Standard. User specifies with "fast", "deep", or tier defaults to Standard.

## World Model Storage

Models are stored at: `~/.claude/MEMORY/RESEARCH/WorldModels/`

| File | Horizon |
|------|---------|
| `INDEX.md` | Summary of all models with last-updated dates |
| `6-month.md` | 6-month outlook |
| `1-year.md` | 1-year outlook |
| `2-year.md` | 2-year outlook |
| `3-year.md` | 3-year outlook |
| `5-year.md` | 5-year outlook |
| `7-year.md` | 7-year outlook |
| `10-year.md` | 10-year outlook |
| `15-year.md` | 15-year outlook |
| `20-year.md` | 20-year outlook |
| `30-year.md` | 30-year outlook |
| `50-year.md` | 50-year outlook |

## Context Files

| File | Purpose |
|------|---------|
| `ModelTemplate.md` | Template structure for world model documents |
| `OutputFormat.md` | Template for TestIdea results output |

## Skill Integrations

This skill orchestrates multiple PAI capabilities:

- **RedTeam** — Adversarial stress testing of ideas against each horizon
- **FirstPrinciples** — Decompose idea assumptions into hard/soft/assumption constraints
- **Council** — Multi-perspective debate on idea viability across horizons
- **Research** — Deep research for model creation and updates

## Voice Notification

Before any workflow execution:
```bash
curl -s -X POST http://localhost:8888/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "Running WORKFLOW_NAME in the World Threat Model Harness", "voice_id": "YOUR_VOICE_ID_HERE"}'
```

## Customization Check

Before execution, check for user customizations at:
`~/.claude/skills/PAI/USER/SKILLCUSTOMIZATIONS/WorldThreatModelHarness/`

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

FAQ

How long do different tiers take?

Fast runs in about 2 minutes, Standard around 10 minutes, and Deep can take up to 1 hour depending on research depth.

Where are the world models stored and how can I view them?

Models are saved under ~/.claude/MEMORY/RESEARCH/WorldModels/ with per-horizon files and an INDEX.md summarizing last-updated dates; use the view models workflow to read summaries.