Worldthreatmodel

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

Tier System

All workflows support three execution tiers:

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

World Model Storage

Models are stored at: $PAI_DIR/MEMORY/RESEARCH/WorldModels/

Horizon Models (base views)

Scenario Models (alternative futures)

Stored at: $PAI_DIR/MEMORY/RESEARCH/WorldModels/Scenarios/

Context Files

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:

curl -s -X POST http://localhost:31337/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "Running WORKFLOW_NAME in the World Threat Model Harness", "voice_id": "fTtv3eikoepIosk8dTZ5"}'

Customization Check

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

Gotchas

• 11 time horizons (6mo-50yr). Don't over-index on short-term predictions — the value is in long-term structural analysis.
• Threat models are hypothetical. Present as scenarios with probability ranges, not predictions.
• Update models when major world events occur. Static threat models decay in accuracy.

Examples

Example 1: Test an investment thesis

User: "threat model my bet on AI-first content creation"
→ Analyzes across 11 time horizons (6mo to 50yr)
→ Identifies structural risks at each horizon
→ Returns probability-weighted scenario matrix

Example 2: Stress test a strategy

User: "what could go wrong with our newsletter business model?"
→ Maps threat vectors: market, technology, regulatory, competitive
→ Returns prioritized risk register with mitigations

Execution Log

After completing any workflow, append a single JSONL entry:

echo '{"ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'","skill":"WorldThreatModel","workflow":"WORKFLOW_USED","input":"8_WORD_SUMMARY","status":"ok|error","duration_s":SECONDS}' >> ~/.claude/PAI/MEMORY/SKILLS/execution.jsonl

Replace WORKFLOW_USED with the workflow executed, 8_WORD_SUMMARY with a brief input description, and SECONDS with approximate wall-clock time. Log status: "error" if the workflow failed.

World Model Template

Use this template when creating or updating any world model document.

Frontmatter

---
horizon: {6-month | 1-year | 2-year | 3-year | 5-year | 7-year | 10-year | 15-year | 20-year | 30-year | 50-year}
last_updated: YYYY-MM-DD
version: {integer, increment on each update}
confidence: {high | medium | low}
confidence_note: {brief explanation of confidence level}
---

Required Sections (all mandatory)

1. Executive Summary (~300 words)

High-level snapshot of the world at this time horizon. What are the 5-7 defining characteristics of this period? What has changed most from today? This section should be readable standalone as a briefing.

2. Geopolitics & Power Dynamics (~600 words)

• Great power competition (US, China, EU, Russia, India, emerging blocs)
• Alliance structures and shifts
• Regional hotspots and conflict trajectories
• Governance models and institutional evolution
• Trade and economic bloc dynamics

3. Technology & AI Trajectory (~800 words)

• AI capability trajectory (reasoning, autonomy, multimodal, embodied)
• Compute and infrastructure trends
• Key technology breakthroughs expected or underway
• Technology adoption curves and societal integration
• Digital infrastructure and platform dynamics
• Cybersecurity landscape evolution

4. Economics & Markets (~600 words)

• Global GDP and growth trajectory
• Dominant economic models and shifts
• Labor market transformation (automation, new roles)
• Financial system evolution (crypto, CBDC, traditional)
• Key sector disruptions and emergences
• Wealth distribution and inequality dynamics

5. Society, Culture & Demographics (~500 words)

• Population dynamics and migration
• Cultural shifts and value changes
• Education transformation
• Media and information ecosystem
• Mental health and wellbeing trends
• Urban vs. rural dynamics

6. Environment & Climate (~500 words)

• Climate trajectory and milestone projections
• Energy transition progress
• Resource scarcity and access
• Adaptation vs. mitigation progress
• Biodiversity and ecosystem changes

7. Security & Conflict (~500 words)

• Kinetic conflict landscape
• Cyber and information warfare
• Biosecurity and pandemic preparedness
• Nuclear and WMD landscape
• Non-state actor evolution
• Critical infrastructure vulnerability

8. Wildcards & Black Swans (~400 words)

• 3-5 low-probability (<10%) but high-impact scenarios
• For each: trigger conditions, probability estimate, impact severity, recovery timeline
• At least one positive wildcard (breakthrough, unexpected cooperation)
• At least one catastrophic wildcard (systemic failure, conflict escalation)

9. Confidence Assessment & Key Uncertainties (~300 words)

• Overall confidence in this model's predictions
• Top 5 uncertainties that could invalidate major predictions
• Which sections have highest/lowest confidence and why
• What new information would most change this model

Writing Guidelines

• Use hedging language appropriately: "likely," "projected," "current trajectory suggests," "if trends continue"
• Cite reasoning: Don't state trends without explaining WHY (demographic data, technological trajectory, historical patterns)
• Be specific: Name countries, companies, technologies, numbers when possible
• Distinguish certainty levels: Some trends are near-certain (demographics); others are speculative (geopolitical realignment)
• Near-term models (6mo-3yr): Heavier on current data extrapolation, specific events
• Mid-term models (5yr-10yr): Balance of trend extrapolation and structural analysis
• Long-term models (15yr-50yr): Focus on structural forces, megatrends, scenario analysis

Minimum Length

Each model should be approximately 4,000-6,000 words (~8-12 pages). No section should be fewer than the word count guidance above. Depth > breadth — substantive analysis over surface coverage.

TestIdea Output Format

When the TestIdea workflow completes, present results in this format.

Header

# 🌍 World Threat Model Analysis

**Idea/Strategy:** {user's input, summarized in 1-2 sentences}
**Tier:** {Fast | Standard | Deep}
**Date:** {YYYY-MM-DD}
**Models Last Updated:** {date of oldest model used}

Executive Verdict

## Executive Verdict

**Overall Assessment:** {1-2 sentence summary of how the idea holds up across all horizons}

**Strongest Period:** {which horizon(s) are most favorable and why, in 1 sentence}
**Greatest Risk Period:** {which horizon(s) are most threatening and why, in 1 sentence}
**Recommendation:** {actionable 1-2 sentence recommendation}

Per-Horizon Analysis

For EACH of the 11 horizons, present a clearly separated section:

## ━━━ {HORIZON} Horizon ━━━

**Verdict:** {🟢 FAVORABLE | 🟡 NEUTRAL | 🔴 UNFAVORABLE}
**Confidence:** {HIGH | MEDIUM | LOW}

### Key Factors
- {Factor 1 from the world model that affects this idea}
- {Factor 2}
- {Factor 3}
- {Factor 4 if applicable}

### Analysis
{2-4 paragraphs of substantive analysis. How does the world state at this horizon
specifically interact with the idea? What thrives? What breaks? What adapts?}

### Assumptions Tested (FirstPrinciples)
- **Hard constraint:** {constraint that cannot change}
- **Soft constraint:** {constraint that could shift}
- **Assumption challenged:** {assumption the idea relies on that may not hold}

### Adversarial Findings (RedTeam)
- **Attack vector:** {strongest argument against the idea at this horizon}
- **Steelman:** {strongest defense of the idea at this horizon}

Synthesis Section (Standard and Deep tiers only)

## Cross-Horizon Synthesis

### Timeline of Viability
{A narrative arc showing how the idea's viability changes over time.
Not just "good then bad" — explain the dynamics, inflection points,
and what drives the transitions.}

### Critical Inflection Points
1. **{Year/Period}:** {What changes and why it matters for this idea}
2. **{Year/Period}:** {Second inflection point}
3. **{Year/Period}:** {Third if applicable}

### Adaptation Recommendations
{If the idea needs to evolve to survive longer horizons, what should change and when?}

Council Debate Summary (Deep tier only)

## Council Deliberation

### Strongest Argument FOR
{Council's strongest bull case across all horizons}

### Strongest Argument AGAINST
{Council's strongest bear case across all horizons}

### Points of Consensus
{Where all perspectives agreed}

### Unresolved Tensions
{Where legitimate disagreement remains}

Footer

---
*Analysis generated by World Threat Model Harness | Models version: {dates} | Tier: {tier}*
*For model updates: "update world models" | For different depth: "test [idea] deep/fast"*

Tier-Specific Inclusion

Test Idea Against World Threat Models

Test any idea, strategy, investment, brand, or concept against all 11 persistent world models to assess viability across time horizons.

When to Use

• User says "test this idea," "how will this hold up," "test my strategy," "stress test this"
• User provides an idea/strategy/investment and wants temporal viability analysis
• User wants to understand when an idea breaks or thrives

Prerequisites

• World models must exist at ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/
• If models don't exist, prompt user to run UpdateModels workflow first

Tier Detection

Detect from user prompt:

• "fast" or "quick" → Fast tier
• "deep" or "thorough" or "comprehensive" → Deep tier
• No modifier → Standard tier (default)

Workflow Steps

Step 0: Validate Models Exist

Check ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/ for all 11 model files.
If any missing: "World models incomplete. Run 'update world models' first."
If models older than 30 days: warn user but proceed.

Step 1: Voice Notification

curl -s -X POST http://localhost:31337/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "Testing your idea against all eleven world threat models at TIER tier", "voice_id": "fTtv3eikoepIosk8dTZ5"}'

Step 2: Extract and Decompose the Idea

Before hitting it with world models, decompose the idea:

1. State the idea in 1-2 sentences 2. Identify core assumptions the idea relies on (market conditions, technology state, cultural norms, regulatory environment, competitive landscape) 3. Identify success dependencies — what must remain true for this to work?

For Standard and Deep tiers: Invoke FirstPrinciples skill to classify assumptions:

• Hard constraints (physics, demographics, math)
• Soft constraints (policy, regulation, cultural norms)
• Assumptions (unvalidated beliefs the idea depends on)

Step 3: Run Against World Models

Read all 11 model files from ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/.

Fast Tier (~2 min)

Single-agent analysis: 1. Read all 11 models sequentially 2. For each horizon, generate: Verdict (🟢/🟡/🔴) + 2-3 bullet points 3. Write Executive Verdict 4. Output using abbreviated format from OutputFormat.md

Standard Tier (~10 min)

Parallel agent analysis: 1. Spawn up to 11 parallel agents (Task tool, run_in_background: true) 2. Each agent:

• Reads ONE world model document
• Analyzes the idea against that specific horizon
• Tests each assumption against the horizon's conditions
• Returns: Verdict, Key Factors, Analysis, Assumptions Tested

3. After all agents return, invoke RedTeam skill with:

• Prompt: "Attack this idea across all time horizons. Here are the per-horizon analyses: {results}"
• Extract adversarial findings per horizon

4. Synthesize Cross-Horizon Synthesis section 5. Output using full format from OutputFormat.md

Deep Tier (up to 1 hr)

Full capability invocation: 1. FirstPrinciples (if not already run): Full deconstruct → challenge → reconstruct cycle on the idea 2. Research update check: For each horizon, run quick Research check for any new developments that affect this specific idea 3. Parallel horizon analysis: Same as Standard but with deeper prompts and longer analysis per horizon 4. RedTeam (32 agents): Full adversarial analysis of the idea across all horizons 5. Council: Multi-agent debate on the idea's long-term viability

• Prompt: "Debate the viability of {idea} across time horizons from 6 months to 50 years. Consider: {per-horizon results}"
• Extract Council Deliberation section

6. Synthesize all findings 7. Output using complete format from OutputFormat.md (all sections)

Step 4: Format Output

Use the template in OutputFormat.md (loaded from skill root). Ensure:

• Each horizon is clearly separated with its own section header
• Verdicts use consistent emoji indicators
• Confidence levels reflect model confidence × analysis certainty
• Adversarial findings attribute to specific horizon contexts

Step 5: Voice Summary

curl -s -X POST http://localhost:31337/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "Analysis complete. SUMMARY_OF_EXECUTIVE_VERDICT", "voice_id": "fTtv3eikoepIosk8dTZ5"}'

Output Format

See OutputFormat.md in the skill root directory.

Integration Points

Error Handling

• If a parallel agent fails: continue with remaining agents, note missing horizon in output
• If a skill invocation fails: degrade gracefully (e.g., skip Council section, note in footer)
• If models are stale (>90 days): prominently warn in header, recommend update

Update World Models

Refresh or create world model documents using deep research and user-provided analysis.

When to Use

• User says "update world models," "refresh models," "new analysis for models"
• User provides new information/analysis to incorporate into models
• Models are stale (>30 days since last update)
• Initial model population (no models exist yet)

Prerequisites

• Model template at ~/.claude/skills/WorldThreatModel/ModelTemplate.md
• Research skill available for web research

Workflow Steps

Step 0: Check Existing State

Read ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/INDEX.md (if exists)
Inventory which models exist and their last_updated dates
Determine: full creation vs. targeted update

Step 1: Voice Notification

curl -s -X POST http://localhost:31337/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "Updating world threat models. This will take several minutes as I research current state for each time horizon.", "voice_id": "fTtv3eikoepIosk8dTZ5"}'

Step 2: Determine Update Scope

Full Creation (no models exist or user says "rebuild all"):

• Create all 11 models from scratch
• Use parallel research agents for efficiency

Targeted Update (models exist, user provides new analysis OR routine refresh):

• If user provided analysis: incorporate into relevant horizons
• If routine refresh: research what's changed since last_updated date
• Update only affected models

Single Horizon (user specifies "update the 5-year model"):

• Research and update only the specified horizon

Step 3: Research Current State

For each model being created or updated:

1. Invoke Research skill (Standard or Extensive mode depending on scope):

• Query: "Current global state and projections for {HORIZON} timeframe: geopolitics, AI/technology trajectory, economics, society, environment, security. Focus on developments since {last_updated or 'baseline'}."
• For near-term (6mo-3yr): Focus on current events, immediate trends, specific forecasts
• For mid-term (5yr-10yr): Focus on trend trajectories, structural shifts, emerging patterns
• For long-term (15yr-50yr): Focus on megatrends, demographic forces, technological paradigm shifts

2. If user provided analysis: Integrate user's insights as primary source, research as supporting

3. Parallel execution: When creating multiple models, spawn parallel agents:

• Near-term batch: 6mo, 1yr, 2yr, 3yr (4 agents)
• Mid-term batch: 5yr, 7yr, 10yr (3 agents)
• Long-term batch: 15yr, 20yr, 30yr, 50yr (4 agents)
• Each agent uses Research skill for its specific horizon

Step 4: Write Model Documents

For each model, following ModelTemplate.md:

1. Write frontmatter (horizon, last_updated: today, version: increment, confidence level) 2. Write all 9 required sections with minimum word counts 3. Ensure hedging language for predictions 4. Include specific data points, named entities, cited reasoning 5. Write Wildcards section with probability estimates

Save to: ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/{horizon}.md

Step 5: Update INDEX

Write/update ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/INDEX.md:

# World Threat Models — Index

Last full update: {date}

| Horizon | File | Last Updated | Version | Confidence |
|---------|------|-------------|---------|------------|
| 6 months | 6-month.md | YYYY-MM-DD | N | high/med/low |
| 1 year | 1-year.md | YYYY-MM-DD | N | high/med/low |
| ... | ... | ... | ... | ... |

## Update History

- YYYY-MM-DD: {what was updated and why}

Step 6: Voice Completion

curl -s -X POST http://localhost:31337/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "World models updated. N horizons refreshed with current research.", "voice_id": "fTtv3eikoepIosk8dTZ5"}'

Agent Prompt Template (for parallel model creation)

When spawning agents for model creation, use this prompt structure:

CONTEXT: You are creating a world threat model for the {HORIZON} time horizon.
The current date is {TODAY}. This model will be used to test ideas, strategies,
and investments against projected future world states.

TASK: Write a comprehensive world model document following this template:
{INSERT ModelTemplate.md CONTENT}

REQUIREMENTS:
- Minimum 4,000 words across all sections
- Use hedging language for predictions ("likely," "projected," "if trends continue")
- Be specific: name countries, companies, technologies, cite numbers where possible
- Near-term models: heavier on current data extrapolation
- Long-term models: focus on structural forces and megatrends
- Include at least 4 wildcards with probability estimates
- Rate your overall confidence and explain why

RESEARCH: Use WebSearch to find current data, forecasts, and analysis relevant to
this {HORIZON} timeframe across all sections (geopolitics, technology, economics,
society, environment, security).

SLA: Complete within 5 minutes.

OUTPUT: Write the complete model document in markdown format following the template exactly.
Include the frontmatter with horizon, last_updated, version: 1, and confidence rating.

Integration Points

State Management (Loop Compatibility)

• Models on disk are the state
• Step 0 reads existing state to determine what needs updating
• Each model is independently updatable
• INDEX.md tracks aggregate state

View World Models

Read and display the current state of world threat models.

When to Use

• User says "view world models," "show models," "current models," "model status"
• User wants to understand what's in the current models
• User wants to check model freshness before running TestIdea

Workflow Steps

Step 1: Voice Notification

curl -s -X POST http://localhost:31337/notify \
  -H "Content-Type: application/json" \
  -d '{"message": "Checking current world model state", "voice_id": "fTtv3eikoepIosk8dTZ5"}'

Step 2: Read INDEX

Read ~/.claude/PAI/MEMORY/RESEARCH/WorldModels/INDEX.md. If it doesn't exist: "No world models found. Run 'update world models' to create them."

Step 3: Determine View Scope

Overview (default — no specific horizon mentioned):

• Display the INDEX table with all horizons, dates, versions, confidence
• For each model, show 2-3 sentence Executive Summary excerpt
• Flag any models older than 30 days as stale

Single Horizon (user says "show me the 5-year model"):

• Read and display the full model document for that horizon
• Include all sections

Comparison (user says "compare near-term vs long-term"):

• Side-by-side key themes from selected horizons
• Highlight where short-term and long-term trends conflict

Step 4: Staleness Check

For each model, compare last_updated to today:

• < 7 days: 🟢 Fresh
• 7-30 days: 🟡 Current
• 30-90 days: 🟠 Aging — recommend refresh
• > 90 days: 🔴 Stale — strongly recommend update

Step 5: Output

# 🌍 World Threat Model Status

| Horizon | Last Updated | Version | Confidence | Freshness |
|---------|-------------|---------|------------|-----------|
| 6 months | YYYY-MM-DD | N | HIGH | 🟢 Fresh |
| 1 year | YYYY-MM-DD | N | MEDIUM | 🟡 Current |
| ... | ... | ... | ... | ... |

## Summaries

### 6-Month Horizon
{2-3 sentence executive summary excerpt}

### 1-Year Horizon
{2-3 sentence executive summary excerpt}

...

## Recommendations
- {Any models needing refresh}
- {Any notable changes since last update}

Integration Points

None — this is a read-only workflow.