.. _promy-reap:

****************************************************
PROMY:REAP --- Harvest and Navigate
****************************************************

**Growth cycle:** Harvest |  **ZION:** Navigating

REAP is the fourth and final stage of the PROMY pipeline. It audits
the full session, closes the books, and produces three tailored
overview pages so each audience gets a landing page calibrated to
their level. After REAP, the model is ready for SISYF.

**Input:** Updated model files from GROW + full session record

**Output:** Three overview pages + final llog entry + closed AA items

**Transition:** Session complete. Model is ready for SISYF.


What REAP Does
================

The prompt has three steps:

**Step 1 --- Full Audit.** Read every file in the model directory
and check: file completeness (all 12 standard files), internal
consistency (predicates/symbols coverage, ``:ref:`` links, ``vvnow``
format), and AA status (resolved/open/new).

**Step 2 --- Three Overview Pages.**

.. list-table::
   :header-rows: 1
   :widths: 25 30 18 20

   * - Page
     - Audience
     - Length
     - Tone
   * - ``overview-beginner.rst``
     - No formal background
     - 800--1200 words
     - Coffee conversation
   * - ``overview-producer.rst``
     - Builders and researchers
     - 1000--1500 words
     - Good README
   * - ``overview-expert.rst``
     - Logicians and mathematicians
     - 1500--2500 words
     - Paper introduction

Each overview must: trace every claim to a formal statement via
``:ref:``, use BEST Names (not Greek letters), note the epistemic
status (e.g., OOv1), and follow all Language Rules from CLAUDE.md.

**Step 3 --- Close the Books.** Finalize ``index.rst`` toctree,
mark AA items resolved, write final llog entries, run ``make html``,
report model-ready-for-SISYF status.


Key Design Choices
====================

- **REAP never modifies formal statements** --- that is GROW's job.
  If REAP finds an error in a formal statement, it flags it as a
  new AA item.
- **Overviews are new content**, not llog extraction --- but they
  must be accurate to the post-GROW model state. Every claim must
  trace to a specific axiom or theorem.
- **Epistemic honesty is mandatory.** At OOv1, nothing is proven.
  Overviews describe what the model *proposes*, not what it
  *establishes*. But they state the proposals clearly, not buried
  under hedging.


Prerequisites
===============

Before running REAP, confirm:

1. **GROW completed and committed.** All repairs applied, all diffs
   human-reviewed, clean git commit.
2. **``make html`` builds cleanly.** Zero warnings from model files.
3. **AA is current.** All resolved items marked, all open items
   have accurate descriptions.
4. **StayC level decided.** Know what epistemic level you are
   assigning (e.g., OOv1) before the overviews are written.
5. **1st-intro.rst reviewed.** The beginner overview complements
   (not duplicates) the existing introduction.
6. **No ``.. keep`` conflicts.** Content marked ``.. keep`` must
   not be overwritten. Overview pages are new files, so low risk,
   but the audit step may surface files with keep markers.


How to Run
============

See :doc:`prompt` for the full generic template with
``{{placeholders}}``. Below is the concrete invocation for e7He
(adapt for other models):

::

  PROMY:REAP e7He

  You are closing the books on the PROMY session for e7He.
  This is a READ-MOSTLY step. You create new overview files and
  audit existing ones. You do NOT modify formal statements (that
  was GROW's job). If you find an error in a formal statement,
  FLAG it --- do not fix it.

  RULES:
  - Do NOT modify axioms.rst, theorems.rst, or any formal statement
  - If you find inconsistencies, report them as new AA items
  - Overview pages are NEW content written by you, not extracted
    from the llog --- but they must be accurate to the model
  - Every claim in an overview must be traceable to a specific
    axiom, theorem, or structural property via :ref: link
  - Follow all Language Rules from CLAUDE.md
  - All claims are at OOv1 per the author's assessment
  - Include standard page prefix/postfix includes per site convention

  MODEL STATE (post-GROW):
  - 14 axioms (7 m0 prerequisites + 7 stage axioms m1--m7)
  - 7 theorems (th1--th7)
  - 3 structural properties (sp1--sp3)
  - 42 predicates
  - Repairs applied: m0.ax3 (maximizes to contributes positively),
    th3 (universal to conditional with Ie variable rename),
    m0.ax5 (OKO to NOT-OK), predicate #19, sp1 justification
  - Deferred: th2 Lyapunov formalization (AA-e7He-Lyapunov-a1)

  EXISTING CONTENT TO COMPLEMENT (not duplicate):
  - 1st-intro.rst has full 7-stage descriptions (~400 lines)
  - silent-corruption/ has public + formal corruption analysis
  - models.rst has submodel architecture table
  - logics.rst has formal disciplines + Iron Maiden test coverage

  STEP 1: FULL AUDIT
  STEP 2: THREE OVERVIEW PAGES
  STEP 3: CLOSE THE BOOKS

  See the REAP prompt template at compiler/promy/reap/prompt.rst
  for the full specification of each step.


.. toctree::
   :maxdepth: 1

   prompt


Completed REAP Runs
======================

- **e7He** (2026-03-30): Full audit (clean), 3 overview pages
  (beginner/producer/expert), books closed. Model ready for SISYF.
  See :doc:`PROMY llog </matheology/hell/ll/promy/b/11/promy_ll_2026m03d29_e7he-pipeline-test>`.