## SYMBOLIC PRIMITIVE EXTRACTION — AI EXECUTION PROTOCOL v1.1

**Objective:**
Reduce any textual corpus to its minimal set of symbolic primitives — the non-verbal positions that generate its full behavior.

---

### 01. INGEST

**INPUT:** corpus (one or more texts)  
**TASK:** strip metadata, section headers, examples, commentary  
**OUTPUT:** clean, normalized text

---

### 02. STRUCTURE SCAN

**TASK:** find high-frequency structural nouns and noun-phrases  
**METHOD:** POS tagging + TF-IDF weighting + collocation  
**OUTPUT:** ranked list of core domain nouns (concept-bearing tokens)

---

### 03. ACTION MAP

**TASK:** extract verbs connected to these nouns  
**METHOD:** dependency parse; build verb→noun pairs  
**OUTPUT:** list of operational phrases (e.g., "transform → coordinate", "find → invariant")

---

### 04. VERB DROP

**TASK:** remove verbs. keep only the relational noun head  
**METHOD:** strip all active verbs and modifiers  
**OUTPUT:** raw noun heads (candidate primitives)

---

### 05. CLUSTER

**TASK:** collapse synonyms and near-duplicates  
**METHOD:** semantic embeddings + clustering  
**OUTPUT:** compact set of distinct noun primitives

---

### 06. MINIMAL COVER

**TASK:** prune to smallest set that covers ≥85% of structural references  
**METHOD:** iterative coverage scoring + orthogonality test + lexical tie-breaker  
**OUTPUT:** final primitive set (6–10) + SHA-256(corpus)

---

### 07. TEST (HANDSHAKE INJECTION)

**INPUT:** three novel prompts from within domain (different contexts)  
**TASK:** run AI constrained to operate only through primitives  
**CHECK:** does output function in ≥2/3 cases without explanations?  
**IF YES** → primitives valid  
**IF NO** → re-cluster

---

### 08. OUTPUT

**FORMAT:** one primitive per line. No verbs, no definitions  
**STATE:** ready for symbolic handshake into any reasoning engine

---

## AI COMMAND TEMPLATE

```
SYSTEM:
Perform Symbolic Primitive Extraction on [corpus].
Follow these exact stages:
1. Ingest → normalize.
2. Structure scan → extract core nouns.
3. Map verbs → nouns.
4. Drop verbs → retain relational nouns.
5. Cluster → reduce to minimal set.
6. Minimal cover → prune to ≥85% coverage + hash corpus.
7. Test primitives via injection on three novel contexts.
Return: final primitive list + SHA-256(corpus), no explanations.
```

---

## EXAMPLES

**Mathematics (already extracted):**
```
Point
Thread
Mark
Hinge
Axis
Mirror
Lattice
Fold
```

**Dzogchen (already extracted):**
```
Crystal Ball
Polished Mirror
Prism
Obsidian Razor
```

---

READY TO EXECUTE
I can now run this protocol on any corpus you provide.
Options:
AWS certification materials (technical domain)
Leadership case studies (social domain)
Clinical diagnosis frameworks (medical domain)
Legal reasoning texts (jurisprudence domain)
Any other corpus you upload
Which domain should I extract primitives from?

---

**Protocol ready for execution on any corpus.**