CoolFace
Apppublic

zkaedi/prime-swarm-hunter

sourceHugging Facemitupdated 7mo agoView on Hugging Face
0likes
App README

🔱 ZKAEDI PRIME — 12-Agent Vulnerability Hunter

Recursively Coupled Hamiltonian Swarm with temporal correlation compound detection.

12 specialized agents navigate an energy landscape built from vulnerability signatures to detect compound vulnerabilities that independent analysis misses.

Architecture

  • —12 Specialized Agents: RECON, SENTINEL, ARCHITECT, ORACLE, HEALER, FORGE, PHANTOM, NEXUS, CIPHER, VANGUARD, ECHO, APEX
  • —PRIME Dynamics: H_t = H_0 + η·H_{t-1}·σ(γ·H_{t-1}) + ε·N(0, 1+β|H_{t-1}|) + κ·Σⱼ C(xᵢ,xⱼ)·Hⱼ
  • —FHN Chaos Engine: Two-field activator/inhibitor prevents premature convergence
  • —Temporal Correlation: Compound detection via historical well-entry event correlation
  • —Vectorized Engine: 1,200+ steps/sec on CPU

Model Tool-Use

python
from gradio_client import Client
import json

client = Client("zkaedi/prime-swarm-hunter")

# Preset scenario
result = json.loads(client.predict("defi_lending_pool", 300, 150, 42, api_name="/api_preset"))

# Custom signatures from your model
signatures = [{"id": "v1", "vuln_type": "reentrancy", ...}]
result = json.loads(client.predict(json.dumps(signatures), 300, 150, 42, api_name="/api_custom"))

print(f"Solo: {result['summary']['solo_detected']}")
print(f"Compound: {result['summary']['compounds_detected']}")
print(f"Risk: {result['summary']['risk_score']}")

Integration Pipeline

Solidity Code → gemma-7b-solidity-energy-signatures → Vulnerability Signatures
                                                            ↓
                                          PRIME Swarm Hunter (this tool)
                                                            ↓
                                                   Compound Detections
                                                            ↓
                              solidity-vuln-auditor-7b → Final Audit Report

Author

ZKAEDI — Offensive Healer | GitHub