aex-assumption-extraction

name	aex - Assumption Extraction
description	Extract hidden assumptions from any content. Surfaces what must be true for claims to hold, enabling deeper analysis.
output	{"format":"prose"}

Assumption Extraction

Input: $ARGUMENTS

Interpretations

Before executing, identify which interpretation matches the user's input:

Interpretation 1 — Find hidden assumptions in a plan or argument: The user has a strategy, proposal, or claim and wants to surface the unstated beliefs it depends on. Interpretation 2 — Uncover personal biases: The user suspects their own thinking is constrained by assumptions they cannot see and wants help identifying their blind spots. Interpretation 3 — Audit what's being taken for granted: The user is looking at a situation, system, or status quo and asking "what is everyone just accepting without questioning?"

If ambiguous, ask: "I can help with extracting assumptions from a specific plan, uncovering your own hidden biases, or auditing what's being taken for granted in a situation — which fits?" If clear from context, proceed with the matching interpretation.

---

Purpose

Every claim, plan, or argument rests on hidden assumptions—things that must be true for the claim to hold. This skill extracts those assumptions, making implicit beliefs explicit.

Why this matters:

• Hidden assumptions are often where claims fail
• Making assumptions explicit enables testing them
• Assumptions reveal worldview and blind spots

---

Assumption Types

Type	Description	Example
Causal	X causes Y	"Marketing -> Sales" assumes causation
Existence	X exists	"The market" assumes a defined market exists
Stability	X won't change	"Current trends continue" assumes stability
Access	We can reach/use X	"Use the API" assumes API access
Capability	We/they can do X	"Team will deliver" assumes capability
Value	X is good/desirable	"Growth is good" assumes growth is valued
Knowledge	We/they know X	"Users understand" assumes knowledge
Resources	X resources available	"Budget exists" assumes funding
Permission	Allowed to do X	"We can change" assumes authority
Timing	X timing is correct	"Now is the time" assumes timing

---

Depth Scaling

Default: 2x. Parse depth from $ARGUMENTS if specified (e.g., "/aex 4x [input]").

Depth	Min Assumptions Found	Min Layers Deep	Min Categories	Min Hidden Assumptions
1x	5	2	2	1
2x	8	3	3	2
4x	12	4	5	4
8x	18	5	7	6
16x	25	6	9	10

These are floors. Go deeper where insight is dense. Compress where it's not.

---

The Extraction Process

Step 1: Identify the Core Claims

Parse the input into distinct claims:

INPUT: [original content]

CORE CLAIMS IDENTIFIED:
1. [Claim 1]
2. [Claim 2]
3. [Claim 3]
...

TOTAL CLAIMS: [N]

---

Step 2: For Each Claim, Extract Assumptions

Apply the assumption extraction questions:

CLAIM: [claim text]

EXTRACTION QUESTIONS:

1. CAUSAL: What must cause what for this to work?
   -> [assumption]

2. EXISTENCE: What must exist that isn't proven?
   -> [assumption]

3. STABILITY: What must stay the same?
   -> [assumption]

4. ACCESS: What must we be able to reach/use?
   -> [assumption]

5. CAPABILITY: What must someone be able to do?
   -> [assumption]

6. VALUE: What must be considered good/desirable?
   -> [assumption]

7. KNOWLEDGE: What must someone know?
   -> [assumption]

8. RESOURCES: What resources must be available?
   -> [assumption]

9. PERMISSION: What must we be allowed to do?
   -> [assumption]

10. TIMING: What timing must be correct?
    -> [assumption]

---

Step 3: Rate Assumption Hiddenness

How obvious is each assumption?

Level	Description	Action
Surface	Explicitly stated or obvious	Note for completeness
Shallow	Easily inferred	Flag for verification
Deep	Non-obvious, requires analysis	Highlight for attention
Buried	Counter-intuitive, easy to miss	Critical to surface

ASSUMPTION HIDDENNESS RATING:

| Assumption | Type | Hiddenness | Risk if Wrong |
|------------|------|------------|---------------|
| [Assumption 1] | Causal | Deep | High |
| [Assumption 2] | Stability | Surface | Low |
| [Assumption 3] | Capability | Buried | Critical |
...

---

Step 4: Identify Assumption Dependencies

Some assumptions depend on others:

ASSUMPTION DEPENDENCIES:

[Assumption A]
    |-- requires [Assumption B]
            |-- requires [Assumption C]

DEPENDENCY CHAINS:
1. [A] -> [B] -> [C] (if C fails, A and B fail)
2. [D] -> [E]

ROOT ASSUMPTIONS (no dependencies):
- [C]
- [E]
- [F]

Note: Root assumptions are most critical to verify

---

Step 5: Categorize by Testability

TESTABILITY ASSESSMENT:

IMMEDIATELY TESTABLE (can verify now):
- [Assumption X]: Test by [method]
- [Assumption Y]: Test by [method]

TESTABLE WITH EFFORT (requires work):
- [Assumption Z]: Would need [what] to test

UNTESTABLE (must accept or reject):
- [Assumption W]: Cannot test because [reason]. Evidence searched: [what you looked for and didn't find]. Closest available proxy: [what partial evidence exists].
  -> Decision: [accept/reject/assume for now]

ALREADY TESTED (evidence exists):
- [Assumption V]: [evidence source]

---

Step 6: Generate Assumption Map

===================================================
ASSUMPTION MAP: [topic]
===================================================

CLAIM 1: [claim text]
|-- [Assumption 1.1] (Type: Causal, Hidden: Deep)
|-- [Assumption 1.2] (Type: Existence, Hidden: Surface)
|-- [Assumption 1.3] (Type: Capability, Hidden: Buried) [!]️ HIGH RISK

CLAIM 2: [claim text]
|-- [Assumption 2.1] (Type: Stability, Hidden: Shallow)
|-- [Assumption 2.2] (Type: Resources, Hidden: Deep)

[Continue for all claims]

===================================================

SUMMARY STATISTICS:
- Total claims analyzed: [N]
- Total assumptions extracted: [M]
- By type: Causal [X], Existence [Y], Stability [Z], ...
- By hiddenness: Surface [A], Shallow [B], Deep [C], Buried [D]
- High-risk assumptions: [count]

===================================================

PRIORITY ASSUMPTIONS (most critical to verify):

1. [Assumption] - Type: [type], Risk: HIGH
   Why critical: [reason]
   How to test: [method]

2. [Assumption] - Type: [type], Risk: HIGH
   Why critical: [reason]
   How to test: [method]

3. [Assumption] - Type: [type], Risk: MEDIUM
   Why critical: [reason]
   How to test: [method]

===================================================

---

Quick Extraction (Abbreviated)

For fast assumption surfacing:

QUICK ASSUMPTION EXTRACTION: [topic]

Claim: [main claim]

Must be true:
[ ] [Causal assumption]
[ ] [Existence assumption]
[ ] [Stability assumption]
[ ] [Capability assumption]
[ ] [Resource assumption]

Most hidden: [which one]
Most risky if wrong: [which one]

---

Example: "We should expand to the European market"

Claims

1. We should expand (action is good)
2. European market is the target (location choice)
3. Expansion is possible (capability exists)

Assumption Extraction

Claim 1: "We should expand"

• CAUSAL: Expansion -> growth/profit
• VALUE: Growth is desirable
• TIMING: Now is the right time
• STABILITY: Current success will continue

Claim 2: "European market"

• EXISTENCE: A viable European market exists for us
• ACCESS: We can enter this market
• KNOWLEDGE: We understand European customers
• PERMISSION: Regulations allow entry

Claim 3: "Expansion is possible"

• CAPABILITY: We can execute international expansion
• RESOURCES: We have capital, people, time
• STABILITY: Our core business won't suffer

Assumption Map Output

PRIORITY ASSUMPTIONS:

1. "We understand European customers" - Knowledge, BURIED
   Why critical: Product-market fit depends on this
   How to test: Customer research, pilot market

2. "Our core business won't suffer" - Stability, DEEP
   Why critical: Expansion could drain resources
   How to test: Capacity analysis, scenario modeling

3. "Regulations allow entry" - Permission, SHALLOW
   Why critical: Could block entire plan
   How to test: Legal review, compliance check

---

Assumption Extraction Patterns

For Plans/Strategies

Focus on: Capability, Resources, Timing, Stability

For Arguments/Claims

Focus on: Causal, Existence, Knowledge, Value

For Predictions

Focus on: Stability, Causal, Timing

For Decisions

Focus on: Value, Access, Permission, Capability

---

Quality Checklist

Before completing:

• All core claims identified
• Each claim analyzed with 10 extraction questions
• Hiddenness rated for each assumption
• Dependencies mapped
• Testability assessed
• Priority assumptions identified
• Assumption map generated

---

Integration

Use with:

• /ai -> Invert assumptions to find blind spots
• /araw -> Test assumptions with Assume Right / Assume Wrong
• /cda -> Find where assumptions differ in other domains