Paper b12 — Extraction Prompt (FORGE Log → Knowledge Base)

Run this prompt in a Claude Code session BEFORE running the b12 writing prompt. It produces a structured knowledge base (KB) that the writing session loads as pre-digested input.

/effort max

TASK: Extract material from 3 FORGE session logs into a structured
knowledge base (KB) for writing Paper b12 (e7Day Model) in the HEAVEN
study series. This is a preparation step — you are NOT writing the
paper itself.

PHASE 1 — ORIENT (read these in order):

1. .claude/CLAUDE.md — especially Core Principle, Language Rules, LLog Rules,
   and the EDEN/BABL/ZION framework
2. The e7Day model source (this is the STRUCTURE you will organize around):
   - source/matheology/model/e7day/overview.rst
   - source/matheology/model/e7day/axioms.rst (21 axioms)
   - source/matheology/model/e7day/theorems.rst (9 theorems)
   - source/matheology/model/e7day/symbols.rst
3. The b12 writing prompt (this tells you what the paper NEEDS):
   source/matheology/hell/ll/study/b/12/b12-prompt-writing.rst
4. Paper b11 (PET) as format reference and for PET↔e7Day connections:
   source/matheology/hell/ll/study/b/11/study_ll_2026m04d03_a1-pet-paper.rst

After reading these, you should know:
- The 8 submodels (m0 VOID through m7 TRUST) and 4 meta-axioms
- The 9 theorems and their dependencies
- What the paper needs to cover (KEY CONTENT section of the writing prompt)
- What format the paper will follow

PHASE 2 — EXTRACT (walk each log systematically):

Read ALL THREE forge session logs, in order:
- source/matheology/hell/ll/forge/b/11/llog.rst (sa1, 273 lines — FORGE infrastructure)
- source/matheology/hell/ll/forge/b/12/llog.rst (sa2, 14K lines — PRIMARY e7Day session)
- source/matheology/hell/ll/forge/b/13/llog.rst (sa3, 10K lines — e7He, but cross-refs e7Day)

For each log, extract ONLY material relevant to paper b12 (e7Day).
Focus on these categories:

A. DESIGN RATIONALES — Why was each axiom formulated this way? What
   alternatives were considered and rejected? (The paper needs to explain
   not just WHAT the axioms say, but WHY they take this form.)

B. TEMPER REFINEMENTS — How did Iron Maiden testing change the
   formulations? What weaknesses were found and how were they addressed?
   (This is the "tested, not just proposed" evidence the paper needs.)

C. REJECTED ALTERNATIVES — Ideas that were explored but deliberately
   not included. These are valuable for the Discussion section ("why
   the model does NOT include X").

D. CONNECTIONS TO PET/JUB — Any material about how e7Day relates to
   PET axioms (paper b11) or JUB axioms (paper b14). Cross-paper
   references the writing session will need.

E. KEY INSIGHTS — Observations, analogies, or explanations that
   illuminate the model but are NOT captured in the formal axiom/theorem
   files. These are often buried in HEAT-phase exploration or
   QUENCH-phase reflections.

F. OSCR/BABL MATERIAL — Anything about the OSCR collapse mechanism,
   BABL origin (th3), or ZION cycle that the paper needs to formally
   introduce (the b12 prompt says: "THIS PAPER INTRODUCES THE BABL/ZION
   FRAMEWORK").

SCOPE RULE: Material that is SOLELY about e7He (hero journey stages,
binary temptation encoding, etc.) belongs to paper b13, not a2. Skip it.
EXCEPTION: If e7He material illuminates an e7Day concept (e.g., how
e7Day's th7 Gate 5 motivated e7He), note it briefly with a pointer.

PHASE 3 — BUILD THE KB:

Save the knowledge base at:
source/matheology/hell/ll/study/b/12/study_ll_2026m04dNN_a2-e7day-extraction-kb.rst
(replace NN with today's date)

Structure the KB as follows:

1. One section per e7Day submodel (m0 through m7) + one for meta-axioms (mc)
   + one for theorems + one for cross-paper connections + one for BABL/ZION
2. Within each section, organize by the categories above (A–F)
3. For each extracted item:
   - State the key point concisely
   - Give the source reference: forge session, round, entry, and approximate
     line number (e.g., "Sa2 R1 HEAT e3, b/12/llog.rst ~line 450")
   - If restating the data is more efficient for the paper writer than
     pointing back, restate it. If pointing is sufficient, point.
     Use your judgment: the goal is to MAXIMIZE the paper writer's
     efficiency.
4. End with a "Notes for Other Papers" section for any important
   cross-cutting observations that belong to a3–a7, not a2.

LLOG: Create an extraction session llog at:
source/matheology/hell/ll/study/b/12/study_ll_2026m04dNN_a2-e7day-extraction-llog.rst
(replace NN with today's date)
Follow LLog Rules in CLAUDE.md: append-only, verbatim prompts.

BUILD CHECK: Run "make dev" after writing to check for RST errors.

PHASE 4 — RETROSPECTIVE (do this LAST, append to the llog):

This is the first extraction session in the series. Papers b13 (e7He) and
b14 (JUB) will need similar extraction sessions later. Your experience here
is valuable for improving those future prompts. At the end of the session,
append a "Process Retrospective" section to the llog answering:

1. WHAT WORKED — Which parts of this prompt were most useful? Which
   categories (A–F) yielded the richest material? Was the 3-phase
   structure (orient → extract → build) the right sequence?
2. WHAT DIDN'T WORK — Were any instructions unclear, redundant, or
   unhelpful? Was the scope rule (a2-only) easy to apply or ambiguous?
   Did you run into context limits?
3. KB STRUCTURE — Was organizing by submodel the right choice? Would
   a different organization (e.g., by category first, then submodel)
   serve the paper writer better?
4. POINTING vs. RESTATING — How often did you choose to restate data
   vs. just pointing? What drove those decisions? What ratio felt right?
5. RECOMMENDATIONS FOR a3/a4 — Specific suggestions for adapting this
   prompt for the e7He extraction (a3, ~10K lines in sa3) and the JUB
   extraction (a4, ~50 llog entries across hell/ll/jub/).

Also save these recommendations separately at:
source/matheology/heaven/study/BB/extraction-lessons-learned-for-next-extractons.rst
so the person preparing a3/b14 extraction prompts can find them easily.

WHAT NOT TO DO:
- Do NOT write the paper — that is a separate session
- Do NOT use "validate/verify" — use "test/check" (Language Rule 4)
- Do NOT use YYYY-MM-DD dates — use YYYYmMMdDD (Language Rule 6)
- Do NOT duplicate the full llog content — extract, summarize, and point
- Do NOT skip any of the 3 forge logs — sa1 has FORGE design context,
  sa3 has e7Day cross-references even though it's primarily e7He