Challenge Genealogy Reasoning

Challenge Genealogy Reasoning — Skill 84

Purpose

Reason about challenge lineage to improve future generation. Track which template/mutation combinations produce the best CDI and use that data to make better decisions.

Genealogy Data Structure

{
  "template": "tmpl-{family}-v{N}",
  "lineage": [
    {
      "instance_id": "BOUTS-2026-XXXX",
      "generation": 1,
      "parent": null,
      "mutations": ["none — original"],
      "cdi": 0.82,
      "attempts": 145,
      "status": "retired",
      "framework": "express",
      "domain": "payment_processing"
    }
  ],
  "analysis": "Text analysis of CDI trends, mutation effectiveness, and recommendations"
}

Genealogy Queries

"Why are recent instances scoring lower?"

• Check: what mutations have been applied? Common factor?
• Check: is a specific framework/domain becoming recognizable?
• Example: "Last 4 instances all used Express. Agents have pattern-matched Express debugging. Switch to Fastify/Hono."

"Which canonical engine produces the best CDI?"

• Compare average CDI across all engines, all instances
• Identify: which engines naturally create discrimination vs which need more work

"Should this template be retired or refreshed?"

• CDI trend across last 10 instances — declining = contamination accumulating
• Mutation space remaining — exhausted meaningful dimensions? → retire
• Structural refresh possible? Change interconnection topology, not just surface

"What makes the best mutations?"

• Track: which mutation types correlate with high CDI
• Semantic mutations (changing bug type) typically produce more CDI than structural mutations (changing file layout)
• The bug matters more than the scaffolding

"Is this a template problem or an instance problem?"

• If one instance has low CDI but siblings are fine → instance problem (retire it, generate new)
• If last 3+ instances all have declining CDI → template problem (need new template)
• If ALL templates in a family are declining → family-level contamination (need new engine variant)

Genealogy-Informed Generation

When generating a new instance, check:

1. What mutations produced the best CDI in this template's history?
2. What mutations produced the worst CDI?
3. What frameworks/domains have been overused?
4. What is the maximum mutation depth before template refresh?

Use this data to select mutations that are statistically more likely to produce high CDI.

Template Lifecycle Stages

New Template → Instance Gen 1 (high CDI expected)
  → Instance Gen 2-5 (mutations, CDI monitored)
  → Instance Gen 6-10 (CDI typically starts declining)
  → Template Refresh (structural change, resets mutation depth)
  → Instance Gen 1-5 of refreshed template
  → ...
  → Template Retirement (mutation space exhausted, pattern too recognizable)
  → New Template for same engine family

Maximum mutation depth before refresh: 10 generations

Integration Points

• Mutation Layer (Skill 52): Genealogy informs mutation selection
• Rebalance Recommendations (Skill 83): Template health assessment uses genealogy
• Contamination Doctrine (Skill 49): Genealogy tracks contamination accumulation
• Admin Diagnostics (Skill 87): Template evolution trends shown in dashboards