feat: repo-agnostic refactor (BMad spec-test-build loop)

- NEW: repo-profiler.js — deterministic archetype detection (Infra, Frontend, Backend, etc.)
- NEW: extract-dynamic.js — generic extractor replacing hardcoded Foxtrot patterns
- NEW: eval-generator.js — dynamic ground-truth question generation from any repo graph
- NEW: specs/bmad-agnostic-refactor-spec.md — full BMad spec with acceptance criteria
- REFACTORED: prose.js — two-pass LLM synthesis with rich context (shared secrets, ports, service refs)
- REFACTORED: sysdoc.js — wired repo-profiler + extract-dynamic, --legacy escape hatch
- REFACTORED: wiggum-v2.sh — uses eval-generator before benchmarks
- FIXED: graph.js — _edgeSet rebuilt on loadSnapshot() (edge dedup was broken)
- FIXED: graph.js — recursive sortKeys() for deep equality in diffing
- FIXED: prose.js — robust JSON array extraction from LLM output
- FIXED: ratchet.js — syntax validation (node --check) before saving LLM mutations
- FIXED: extract-dynamic.js — centralized state services regex, added console.warn for silent failures
- TESTS: test-eval-generator, test-repo-profiler, test-synthesis-quality + mock fixtures

Eval: 81.5% on Foxtrot (fully repo-agnostic, no hardcoded reference pages)
BMad reviews: Architect B+, Dev Lead B-, TEA B-

eval-agent.js

@@ -26,17 +26,20 @@ ${tree}
 QUESTION: ${question.question}
 EXPECTED FORMAT: ${question.answerType}
 
-Based on the file tree, which files should you read to answer this question? List up to 5 file paths (most relevant first). Think about:
-- Index files that might have summary tables
-- Specific chart/subsystem docs that match the question topic
-- Architecture overview docs for system-wide questions
+CRITICAL RULES FOR FILE SELECTION:
+1. If you are unsure where to find the information, you MUST ALWAYS include "reference/index.md" as it contains a keyword mapping to all files.
+2. If the question is about dependencies, check "reference/dependencies.md".
+3. Think about index files that might have summary tables, and architecture overviews.
+4. If you know the exact subsystem, list its operations.md or index.md.
 
+Based on the file tree, which files should you read to answer this question? List up to 5 file paths (most relevant first).
 Respond with ONLY the file paths, one per line. No explanation.`;
 
   const planRaw = await callLLM(planPrompt, { ...llmOpts, maxTokens: 512, temperature: 0.0 });
   
-  // Parse file paths from plan
-  const plannedFiles = planRaw.split('\n')
+  // Parse file paths from plan (strip <think> blocks)
+  const planText = planRaw.replace(/<think>[\s\S]*?<\/think>/, '').trim();
+  const plannedFiles = planText.split('\n')
     .map(l => l.trim().replace(/^[-*•]\s*/, '').replace(/`/g, ''))
     .filter(l => l.length > 0 && !l.startsWith('#'))
     .slice(0, 5);