Tech Inference Agent

Tech Inference Agent

Purpose

Phase 3 orchestrator responsible for inferring technologies across all stack layers by analyzing collected signals against detection patterns.

Responsibilities

1. Frontend Inference: React, Angular, Vue, jQuery, Bootstrap, Tailwind, etc.
2. Backend Inference: Web servers, languages, frameworks, databases, CMS
3. Cloud Infrastructure: AWS, Azure, GCP, PaaS platforms (Heroku, Vercel, etc.)
4. CDN/WAF Detection: Cloudflare, Akamai, Fastly, AWS CloudFront
5. Security Posture: CSP, HSTS, WAF presence, security.txt, bug bounty
6. DevOps Detection: CI/CD tools, Docker, Kubernetes signals
7. Third-Party Services: Payments, analytics, auth, CRM, support widgets

Skills Orchestrated

Execute in parallel:

• frontend_inferencer
• backend_inferencer
• cloud_infra_detector
• cdn_waf_fingerprinter
• security_posture_analyzer
• devops_detector
• third_party_detector

Input

Raw signals from Phase 2:

{
  "signals": {
    "http_signals": [...],
    "dns_signals": {...},
    "tls_signals": [...],
    "javascript_signals": [...],
    "html_signals": [...],
    "repository_signals": {...},
    "job_signals": {...},
    "archive_signals": {...}
  }
}

Output

Inferred technologies JSON by category:

{
  "phase": 3,
  "company": "string",
  "inferred_technologies": {
    "frontend": [
      {
        "name": "React",
        "category": "JavaScript Framework",
        "version": "18.x (estimated)",
        "signals": [
          {
            "type": "javascript_global",
            "value": "window.React detected",
            "source": "https://example.com",
            "weight": 30
          },
          {
            "type": "dom_attribute",
            "value": "data-reactroot found",
            "source": "https://example.com",
            "weight": 25
          }
        ],
        "total_weight": 55
      }
    ],
    "backend": [
      {
        "name": "Node.js",
        "category": "Runtime",
        "signals": [...],
        "total_weight": 65
      },
      {
        "name": "Express.js",
        "category": "Web Framework",
        "signals": [...],
        "total_weight": 45
      }
    ],
    "infrastructure": [
      {
        "name": "AWS",
        "category": "Cloud Provider",
        "services_detected": ["CloudFront", "S3", "Route53"],
        "signals": [...],
        "total_weight": 80
      }
    ],
    "security": [
      {
        "name": "Cloudflare WAF",
        "category": "Web Application Firewall",
        "signals": [...],
        "total_weight": 90
      }
    ],
    "devops": [
      {
        "name": "GitHub Actions",
        "category": "CI/CD",
        "signals": [...],
        "total_weight": 40
      }
    ],
    "third_party": [
      {
        "name": "Stripe",
        "category": "Payment Processing",
        "signals": [...],
        "total_weight": 35
      },
      {
        "name": "Google Analytics",
        "category": "Analytics",
        "signals": [...],
        "total_weight": 50
      }
    ]
  },
  "timestamp": "ISO-8601"
}

Pattern Matching

Uses detection patterns from patterns/ directory:

• frontend_patterns.json - Framework detection rules
• backend_patterns.json - Server/language detection
• cloud_patterns.json - Cloud provider signatures
• cdn_waf_patterns.json - CDN/WAF fingerprints
• third_party_patterns.json - Third-party service patterns

Signal Weight Reference

Execution Flow

INPUT: Raw Signals
         │
         ├──────────────────────────────────────────┐
         │              ALL PARALLEL                │
         ├─► frontend_inferencer ─────────────────► │
         ├─► backend_inferencer ──────────────────► │
         ├─► cloud_infra_detector ────────────────► │
         ├─► cdn_waf_fingerprinter ───────────────► │
         ├─► security_posture_analyzer ───────────► │
         ├─► devops_detector ─────────────────────► │
         └─► third_party_detector ────────────────► │
                                                    │
                       MERGE                        │
                         │                          │
                         ▼
         OUTPUT: Inferred Technologies JSON

Error Handling

• Continue inference if some pattern files are missing
• Log unmatched signals for pattern improvement
• Return partial results on inference failures
• Flag low-confidence inferences for manual review