doctrine-corpus: A judgment-eliciting Q&A corpus across the shimo4228 research program

A bilingual (EN + JA) Q&A corpus encoding the documented judgment of four research lines — Agent Knowledge Cycle, Contemplative Agent, Agent Attribution Practice, and Authorship Strategy — plus published articles, as judgment-eliciting pairs designed for LLM-mediated diffusion. v0.1.0 contains 851 examples (334 English / 517 Japanese; 343 judgment / 371 explain / 136 definition / 1 contrast) split 766/85 (seed=42) with per-example metadata recording source line, source path, language, and shape. The corpus is the operational form of Authorship Strategy Layer 4 tactic 7 (LLM-first ingest). It is base-model-independent and is intended as an ingredient for larger instruction-tuning mixes, RAG retrieval, and human reading rather than as a standalone fine-tuning package. A Stage D verification LoRA (Qwen3-8B-4bit, 300 iter on the assembled corpus) was trained as a disposable probe and returned a preliminary FAIL verdict on the judgment-transfer rubric (the corpus retains its deliverable status; the LoRA was never a deposit target). See ADR-0001 for the corpus-as-primary-artifact policy and ADR-0005 for the verification verdict and v0.2.0 hypotheses (data scale, shape distribution, base model effect).