.. _promy-prompt-reap: ********************************************************************* PROMY:REAP --- Generic Prompt Template ********************************************************************* :Created: 2026-03-30 :Origin: Drafted after first PROMY:GROW run on e7He :Status: Draft --- untested When to Use ============= Use this prompt after PROMY:GROW has integrated all HELL evidence and the human has reviewed and approved all diffs. REAP is the final stage: it audits the session, closes the books, and produces three tailored overview pages so the model is ready for SISYF. **Prerequisite check before running REAP:** - GROW completed (all repairs applied and committed)? - Human has reviewed all GROW diffs? - ``make html`` builds cleanly? - No unresolved blockers that would invalidate the overviews? The Prompt ============ Copy and adapt the following. Replace all ``{{placeholders}}``. :: PROMY:REAP {{MODEL_NAME}} You are closing the books on the PROMY session for {{MODEL_NAME}}. 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 STEP 1: FULL AUDIT ==================== Read every file in ``model/{{MODEL_DIR}}/`` and check: **1a. File completeness (Standard Model File Set):** Does the directory contain all required files? :: index.rst — exists? toctree complete? axioms.rst — exists? all axioms present? theorems.rst — exists? all theorems present? symbols.rst — exists? all symbols documented? predicates.rst — exists? all predicates documented? models.rst — exists? submodel table complete? logics.rst — exists? all disciplines listed? aa.rst — exists? all items current? llog.rst — exists? all sessions linked? overview-beginner.rst — TO BE CREATED (this session) overview-producer.rst — TO BE CREATED (this session) overview-expert.rst — TO BE CREATED (this session) **1b. Internal consistency:** - Every predicate used in a math block appears in predicates.rst - Every symbol used in a math block appears in symbols.rst - Every :ref: label resolves (no broken cross-references) - All ``.. vvnow`` annotations use no colon (not ``:`` or ``::``) - Prose descriptions match their math blocks (e.g., if a math block was repaired in GROW, does the prose still match?) **1c. AA status review:** For each item in aa.rst: - If resolved: verify the resolution is correct and mark clearly - If open: verify the description is still accurate and the priority still makes sense given post-GROW state - If new issues found during audit: add them Report the full audit results before proceeding. **1d. Store the full audit:** Write the COMPLETE audit (every predicate checked, every symbol checked, every label listed, every AA item reviewed) to:: hell/llog/promy/audit/audit_{{MODEL}}_reap_{{DATE}}.rst Mark as ``:orphan:``. This file is the regression baseline: the next REAP audit can be diffed against it to show exactly what changed. Include the full tables, not just summaries. Reference this file from the PROMY llog. This full-audit storage is active while the PROMY compiler is stabilizing. It may be switched to summary-only once the process is mature and checks are automated. STEP 2: OVERVIEW PAGES ======================== Create three overview pages. Each serves a different audience. All three must be: - **Accurate:** every claim traces to a formal statement - **Current:** reflects post-GROW state (not the original llog) - **Self-contained:** readable without having read the formal files - **Linked:** uses :ref: and :doc: to connect to formal files for readers who want depth **Important constraints:** - Use BEST Names (from symbols.rst) in prose, not Greek letters - Never use bare "Jubilee" as a standalone noun - Never use "validate/verify" for assessment results - Never claim unproven superlatives with "the" - Include StayC level (OOv1) to signal epistemic status - Each overview should note that the model is at OOv1 and what that means (auditor's initial assessment, not yet independently tested) 2a. overview-beginner.rst ^^^^^^^^^^^^^^^^^^^^^^^^^^^ **Audience:** Someone with no formal background. A curious reader, a potential supporter, someone who heard about the project. **Length:** 800--1200 words. **Structure:** 1. **What is this?** One paragraph. What does the e7He model describe? (The hero journey as a mathematical structure.) 2. **Why does it matter?** Two paragraphs. What problem does it address? (Everyone faces BABL temptations; the model maps them systematically.) What does it predict? (Completing the cycle with integrity produces growth; stopping produces decay.) 3. **The seven stages in plain language.** A brief description of each stage (2--3 sentences each). Use the 1st-intro descriptions as a reference but write fresh prose --- do NOT copy from 1st-intro.rst. The beginner overview should be simpler and shorter than the 1st-intro. 4. **The key insight.** One paragraph. What makes this model different from other hero journey frameworks? (The binary encoding, the BABL mechanism, the mathematical testability.) 5. **Where to go next.** Links to: 1st-intro (fuller descriptions), the silent-corruption piece (a concrete example), and axioms/theorems (for those ready for formalism). 6. **Epistemic status.** One paragraph. This model is at OOv1 --- what does that mean? (It has been formulated, extracted, and checked, but not independently tested. It is a proposal, not a proven result.) **Tone:** Warm, clear, no jargon. Like explaining to a thoughtful friend over coffee. 2b. overview-producer.rst ^^^^^^^^^^^^^^^^^^^^^^^^^^^ **Audience:** Someone building on or with the model. A researcher, a developer, a student who wants to use the model in their own work. **Length:** 1000--1500 words. **Structure:** 1. **Model architecture.** The m0 prerequisite + m1--m7 stage structure. How the binary encoding works. What the coalgebraic formalization buys you (non-termination, productivity, bisimulation). 2. **Key interfaces.** What does the model take as input? (Agent h, FATE, GOAL.) What does it produce? (DESTINY, Ie growth, BABL resistance record.) What are the free parameters? (theta for influence threshold, the specific behavioral criteria for proto-formal predicates.) 3. **Dependencies.** What does e7He inherit from PET and JUB? (Agent set H, God-World containment, domain structure.) What does it add? (Coalgebra, dynamics, game theory.) Cross-model links to e7Tr (roles) and e7Ch (church lifecycle). 4. **Open edges.** Where is the model incomplete? Summarize the key AA items: Lyapunov formalization (th2), CTMC precision (th5), lognormal evidence (sp3), proto-formal predicates. These are the places where a producer can contribute. 5. **How to extend.** How would someone add a new theorem? A new predicate? A new stage axiom? Point to the FORGE and PROMY processes. Note the MODEL_CHANGE convention (once implemented). 6. **Epistemic status.** StayC is OOv1 across all claims. What has been tested (Iron Maiden, 10 tests, all 24 statements). What has NOT been tested (independent replication, empirical grounding of proto-formal predicates). **Tone:** Technical but accessible. Like a good README for a research codebase. 2c. overview-expert.rst ^^^^^^^^^^^^^^^^^^^^^^^^^ **Audience:** A formal logician, mathematician, or philosopher of science who can evaluate the axioms on their own terms. **Length:** 1500--2500 words. **Structure:** 1. **Axiomatic summary.** The 14 axioms grouped by function: framework (m0.ax1--ax7), stage definitions (m1--m7). State each axiom in one line of semi-formal notation (using BEST Names, not just symbols). Reference the formal math blocks via :ref:. 2. **Theorem summary.** The 7 theorems with their logical dependencies. Which axioms does each theorem depend on? Draw the dependency graph in text form. Note which theorems are derivable (th1, th3, th4) vs. which are structural claims (th2, th5) vs. which are game-theoretic arguments (th6, th7). 3. **Structural properties.** sp1 (completeness), sp2 (midpoint maximality), sp3 (lognormal). How they constrain the model. The sp1 ordering justification (Hamming-3, escalation, minimal description length). 4. **Formal disciplines.** What logics are used and why. The S5 + CEM + FOL base. The coalgebraic extension. The dynamical systems layer. The game-theoretic layer. Reference logics.rst. 5. **Known weaknesses and open problems.** Be honest. The Lyapunov formalization gap (th2). The CTMC precision gap (th5). The proto-formal predicate gap. The lognormal evidence gap (sp3). The untested theorem pair (th6+th7). These are not flaws --- they are the model's research frontier. 6. **Relationship to existing literature.** Where does e7He connect to known results? Campbell's monomyth (structural parallel but not derivation). Schelling commitment theory (th6). Arrow impossibility (dissolved by m0.ax3 repair). Lyapunov stability theory (th2 sketch). CTMC absorbing states (th5). Spence signaling (th6 Case 3). 7. **Epistemic status.** OOv1. The model has been through one full FORGE TEMPER cycle (3 rounds, 10 Iron Maiden tests, 24 statements). The auditor (Claude Opus 4.6) proposes PP for 21/24 claims. The author assigns OOv1 to all claims. What would it take to advance? Independent TEMPER round, empirical testing, formal proof of derivable theorems. **Tone:** Dense, precise, no hand-holding. Like a paper introduction for a specialist audience. STEP 3: CLOSE THE BOOKS ========================== **3a. Finalize index.rst:** Add the three overview pages to the toctree. Suggested order:: 1st-intro overview-beginner overview-producer overview-expert axioms theorems models logics predicates symbols silent-corruption/index aa llog **3b. Finalize aa.rst:** - Mark AA-e7He-Overview-a1 as resolved (overviews written) - Add any new items discovered during the audit - Ensure every open item has an accurate description and priority **3c. Finalize llog.rst:** Add the REAP entry with: - Date - What was audited - What was created (three overview pages) - Any inconsistencies found - Final AA status (how many resolved, how many open, how many new) **3d. Append to PROMY llog in HELL:** Append a PROMY:REAP section to ``{{PROMY_LLOG_PATH}}`` recording: - Audit results (file completeness, consistency checks) - Overview pages created (with brief description of each) - AA items resolved and created - Final model status **3e. Update GROW index:** Confirm the completed GROW run entry is accurate. **3f. Update REAP index:** Add the completed REAP run entry. **3g. Run final build:** ``make html`` must succeed with no new warnings from model files. **3h. Report:** Present the final summary: - Files created (3 overview pages) - Files modified (index.rst, aa.rst, llog.rst, PROMY llog) - Audit findings (inconsistencies, new AA items) - Model status: ready for SISYF? (yes/no, with reason) The PROMY session for {{MODEL_NAME}} is now complete. How to Prepare (Before Running the Prompt) ============================================ 1. **Review all GROW diffs.** REAP assumes GROW is approved. If you have doubts about any GROW change, resolve them first. 2. **Read the 1st-intro.** The beginner overview should complement (not duplicate) the existing introduction. Know what's already been said. 3. **Check the existing index.rst.** Know what's already in the toctree so the overview pages slot in cleanly. 4. **Decide the StayC level.** REAP will document the epistemic status. Know what level you're assigning before the overviews are written. (For e7He: OOv1 across all claims.) 5. **Note any ``.. keep::`` markers.** Content marked with ``.. keep`` must not be overwritten or removed. The overview pages are new files, so this is less of a risk than in GROW, but the audit step may surface files with keep markers. Common Pitfalls (anticipated) ================================ 1. **Overviews that claim more than the model proves.** At OOv1, nothing is proven. The overviews must describe what the model *proposes*, not what it *establishes*. Watch for language that implies certainty: "the model shows," "this proves," "we have demonstrated." Use instead: "the model proposes," "this predicts," "the axiom states." 2. **Overviews that are too cautious to be useful.** The opposite error. If every sentence is hedged, the reader learns nothing. State the model's claims clearly, then note the epistemic status separately. "The model proposes X [clear statement]. This is at OOv1 [epistemic note]." 3. **Missing :ref: links.** Every formal claim in an overview should trace to a specific axiom or theorem. If you cannot link a claim, either the claim is wrong or the model is incomplete --- flag it. 4. **Stale AA items.** GROW may have resolved items that aa.rst still lists as open. The audit step catches this. 5. **Check vvnow format.** The correct format is ``.. vvnow `` (no colon = RST comment). If any file uses ``.. vvnow:`` or ``.. vvnow::``, drop the colon(s).