second-order-effects（二阶效应）

Setup

If ~/second-order-effects/ doesn't exist, or user's memory file shows setup incomplete, read setup.md first.

When to Use

User faces a decision with non-obvious downstream effects. Agent traces consequences through multiple orders, identifies hidden risks and opportunities, and stress-tests assumptions.

Architecture

Memory lives in ~/second-order-effects/. See memory-template.md for structure.

~/second-order-effects/
├── memory.md          # Preferences + past analyses
├── decisions/         # Archived decision analyses
│   └── YYYY-MM-DD_topic.md
└── patterns.md        # Learned consequence patterns

Quick Reference

Core Rules

1. Always Go Three Levels Deep

First-order: What happens immediately? Second-order: What does that cause? Third-order: What does THAT cause?

Most people stop at first-order. Competitive advantage lives in second and third.

2. Consider All Stakeholders

Map who is affected at each order:

• Direct participants
• Indirect observers
• Market/ecosystem
• Future self

Each stakeholder creates new consequence chains.

3. Invert the Question

After mapping positive outcomes, ask: "What could go wrong at each level?"

4. Time-Weight Consequences

Near-term consequences feel larger than they are. Apply discount:

• 1st order (now): weight 0.5
• 2nd order (weeks/months): weight 1.0
• 3rd order (years): weight 1.5

Decisions that sacrifice 2nd/3rd order for 1st are usually wrong.

5. Document Predictions

Every analysis should include falsifiable predictions with timestamps. Review quarterly. Update patterns.md when patterns emerge.

Consequence Chain Format

Use this structure for every analysis:

## Decision: [One sentence]

### First Order (Immediate)
- Effect 1 → leads to...
- Effect 2 → leads to...

### Second Order (Weeks-Months)
- [Effect 1] causes → ...
- [Effect 2] causes → ...

### Third Order (Months-Years)
- [Second-order effect] causes → ...

### Stakeholder Map
| Who | 1st Order | 2nd Order | 3rd Order |
|-----|-----------|-----------|-----------|

### Inversion (What Could Go Wrong)
- Risk at 2nd order: ...
- Risk at 3rd order: ...

### Decision: [Proceed/Pause/Reject] because [reason tied to 2nd/3rd order]

Common Traps

• Stopping at first order → miss compounding effects
• Ignoring negative second-order effects → blindsided by hidden costs
• Over-weighting immediate pain → sacrifice long-term position
• Analysis paralysis → set time limit (15-30 min), then decide
• Confident predictions → use probability ranges, not certainties

Scope

This skill ONLY:

• Analyzes decisions using consequence chains
• Stores analyses in ~/second-order-effects/
• Learns patterns from past decisions

This skill NEVER:

• Makes decisions for the user
• Accesses external data without request
• Modifies its own SKILL.md

Security & Privacy

Data that stays local:

• Decision analyses in ~/second-order-effects/
• Learned patterns and preferences

This skill does NOT:

• Send data externally
• Access files outside its directory
• Make network requests

Related Skills

Install with clawhub install <slug> if user confirms:

• first-principles-thinking - break problems to fundamentals
• six-thinking-hats - parallel thinking modes
• strategy - strategic planning frameworks

Feedback

• If useful: clawhub star second-order-effects
• Stay updated: clawhub sync