HISTORICAL CHANGE LOGS AND MAINTENANCE NOTES — PRESERVED BODY

STATUS
This material is preserved from the legacy prompt-suite page. It documents how the suite's translation-review and terminology safeguards evolved over time. It should be treated as a maintenance archive and rule-source inventory, not as a separate prompt to run unless explicitly adapted.

CURRENT GOVERNING INTERPRETATION — Batch 12

1. Historical changelog details are retained because they explain why newer safeguards exist.
2. When a historical rule conflicts with a newer prompt body or the Global Kernel, follow the newer, more specific rule unless the user explicitly asks to restore the older behavior.
3. Do not delete historical notes merely because their version numbers are old; they may contain specific examples and rationale that remain useful.
4. Do not copy old terminology rules blindly into new prompts. Promote them only through context-aware checks.
5. Keep the user's corrections already applied in this suite: equanimity is context-dependent, and the retired one-off Self-stage row remains retired.
6. Preserve this archive as a reference for future Protocol B maintenance and for diagnosing prompt drift.

PRESERVED LEGACY CHANGELOG / MAINTENANCE BODY

Changelog: Prompt 6 (v4.1/4.2 → v5.2)

A. Scope & Philosophy

NEW — Universal scopeThe former Chan/Mantrayāna Domain Pack has been generalized. v5.2 applies core protections to all Buddhist translations (Theravāda, Mahāyāna, Vajrayāna), not just Chan/Mantrayāna subsets.

STRENGTHENED — Completeness & standardsv5.2 keeps the ISO-17100 bilingual revision pass and “Completeness Above All” gate and foregrounds them as a hard precondition for any fluency edits.

STRENGTHENED — Scope LockRe-asserts “no cross-doc bleed” unless extra sources are explicitly provided.

B. Terminology & Style Controls

NEW — Global Buddhist Terminology Guard (Loanword-First Rule)Technical Buddhist lexemes now default to the original loanword (IAST for Sanskrit/Pāli; Wylie for Tibetan) with one-time bracketed English gloss. After first mention, the loanword persists (no swapping back to English).Examples enforced: prajñā [wisdom] → thereafter prajñā; śūnyatā, anātman, tathatā, upāya, bhūmi, prajñāpāramitā; ālayavijñāna; trikāya must list dharmakāya / sambhogakāya / nirmāṇakāya (first-mention gloss only).

NEW — No-paraphrase enforcementFor technical terms, v5.2 explicitly forbids smoothing into English paraphrases (e.g., “native luminosity,” “true self”) unless used as a bracketed gloss after the loanword.

NEW — Concordance lockA mini-termbase must be built and a concordance pass run. v5.2 directs reviewers to flag calques/synonym drift (e.g., prajñā → “insight” later) as Terminology errors (MQM).

NEW — Formatting rules for loanwordsUse full diacritics and italicize loanwords on first occurrence per house style; keep diacritics thereafter. If a platform cannot render diacritics, note the fallback (“prajna”) in the report but keep IAST locked in the term list.

NEW — Narrow “conventional English” exceptionsWhen there is a deeply entrenched English title/rubric, retain the English running form and give the original at first mention, e.g., Heart Sūtra (Prajñāpāramitā-hṛdaya), Four Noble Truths (cattāri ariyasaccāni), Dependent Origination (pratītyasamutpāda).Within doctrinal prose, still discuss the lexeme using the loanword after its first gloss.

EXPANDED — Transliteration policyv5.2 clarifies IAST for Sanskrit/Pāli and Wylie for Tibetan on first mention, with optional Tibetan script if the audience warrants it. Chinese can be included at first mention when helpful.

KEPT/CLARIFIED — “Self vs Self” policyThe nuanced anātman guidance remains and is more clearly tied to doctrine identification before choosing renderings (e.g., Self ≠ zhēn wǒ in Buddhist contexts).

KEPT — Awareness vs MindfulnessRetained with explicit language pairs and examples.

REFERENCE GLOSSARIES — GuardrailsThe reference CN/Tib glossaries are kept but v5.2 emphasizes use only when relevant to the current language pair and do not override the Loanword-First rule for technical lexemes.

C. Structural Fidelity & Non-compression

KEPT/REFOCUSED — Structure & Compression Locksv5.2 preserves: 1:1 Q/A lines, full doctrinal chains, and the rule to never summarize lines naming consciousnesses, bodies, seals, mantras, or scripture citations.

KEPT — Variant handlingPreference for literal translation + [bracketed gloss] over synonym substitution is reaffirmed; do not delete the literal.

D. Execution & QA Workflow

KEPT/SHARPENED — Step 0 Coverage & Structure GateMust pass: 1:1 segmentation, structural mirror, punctuation/brackets parity, keyword salience sweep, literal-flag checks (e.g., 不住/无相/不可得/化空/三界/空乐明), mantras/diacritics preserved, length-delta flags, and targeted back-translations.

KEPT — MQM Accuracy ScanPer-segment Omission/Addition/Mistranslation/Terminology tagging remains; now tightly coupled to the concordance lock.

KEPT — Doctrine-Sensitive Triggers (now global, not Chan-only)
Lights/visions: insert “no attachment to marks (signs)” once; never valorize visions.

ālayavijñāna: explicitly note “still a consciousness; not the goal—must open it and see the nature.”

Trikāya: list all three bodies; no partials or euphemisms.

Powers (神通): enforce sequencing (道通 / 漏尽通 first) and “powers are not to be sought.”

Mark-avoidance & non-abiding: injunction diction must not be softened.

KEPT — Reporting packCoverage Report, MQM Summary, Risk Register, Style Conventions, Diacritics/Tibetan check, Back-translation note, and the structured list of Critical/Terminology/Style/Minor issues all remain.

KEPT — Staged delivery“Approve clean copy” gate, chunk size cap, and Safety Threshold (pause if 3 consecutive segments show >25% delta) retained.

E. What changed specifically vs. v4.1/4.2

Replaced / Generalized
The Chan/Mantrayāna Domain Pack as a special case → Universal Buddhist Terminology Guard (applies to all Buddhist texts). The domain-specific term-lock list is no longer the only protective mechanism; the loanword-first rule now serves as the primary global guard.

Added
Loanword-First Rule, Concordance lock, Formatting/diacritics/italicization rules, Entrenched English exceptions, Platform fallback note, and a stricter no-paraphrase policy for technical lexemes.

Strengthened
Scope Lock, ISO-17100 gate, literal-flag doctrine checks, and doctrine-trigger rules now clearly apply across traditions, not just Chan contexts.

Clarified
When to keep English rubrics (titles/headings) vs. when to persist with loanwords inside prose.

That bracketed gloss is the only permitted way to help fluency for technical terms; do not swap out the loanword thereafter.

Unchanged (but reaffirmed)
MQM framework, 1:1 lineation, Q/A preservation, mantras/diacritics verbatim, back-translation sampling, staged delivery, and safety thresholds.

F. Practical Impact (why these changes)

Prevents terminology drift (e.g., prajñā → “insight,” śūnyatā → “void”) by locking the canonical term and surfacing glosses only once.

Ends “creative smoothing” of doctrinal terms via the no-paraphrase rule; fluency is achieved with bracketed glosses, not substitutions.

Raises cross-tradition consistency by applying the doctrine triggers (ālayavijñāna, trikāya, powers, non-abiding) universally.

Stops silent calquing through the concordance pass and MQM Terminology flags.

Preserves philology + readability via strict diacritics/italic guidelines and narrowly-defined English exceptions.

Reduces structural loss by re-emphasizing the Step-0 gate and compression locks before any stylistic polish.

Prompt 6 – Change Log
v5.2 → v5.3
Focus: close completeness gaps; enforce quote/contrast coverage; add ratio sentinels; add read-scan.
Step 0 strengthened (Coverage & Completeness Gate)

Sentence/paragraph alignment mandate (e.g., Gale–Church or equivalent) with strict 1:1 segment counts; any unaligned source/target segment ⇒ Gate = FAIL until fixed.

Length-ratio sentinel per aligned pair (initial band 0.6–1.6 for ZH↔EN; tunable). Three consecutive out-of-band flags ⇒ STOP & FAIL.

Monolingual Read-Scan + Bilingual Spot-Audit added: quick target read for logic jumps + bilingual skim to confirm every sentence/critical clause is covered. Any untraceable jump ⇒ FAIL.

Quoted-string extraction & contrastive patterns check (NEW): extract all substrings in 「」/『』/“ ” and patterns like X就是X，不是Y; build a source→target mini-table; any missing item ⇒ FAIL.

Punctuation/marker parity tightened: counts must match for quotes, brackets, em dashes, numerals/dates, scripture titles, mantras; any delta requires a “Deviation: Element=…, Reason=…, Evidence=…” note.

Keyword & Entity Presence Sweep clarified

Keep exactly 25 items (≥10 doctrinal, ≥5 proper names/titles, ≥3 numerals/dates).

If any are missing in the target, propose exact insertions or give a justification (operationalizes MQM “Omission”).

Literal doctrinal flags reiterated (e.g., 不住→do not dwell; 无相→signless; 不可得→unobtainable; 化空→dissolve into emptiness; 三界→Desire–Form–Formless; 空乐明→emptiness, bliss, and clarity) and mantras/diacritics must be verbatim.

Length-delta scan formalized: any segment >±25%; three consecutive breaches ⇒ FAIL, correct, re-run Step 0.

Back-translation spot check introduced: pick 3–5 densest lines (by doctrinal-term density), note SegIDs and parity verdicts.

Reporting: Certificate Line expanded

Add fields for structural counts and length-delta stats; require MQM(O/A/M/T) tallies.

Formatting guardrails (CommonMark) emphasized

Escape numbered items (1\.) to prevent auto-lists.

Use plain-text code fences (```text) for CSV or literal blocks.

v5.3 → v5.4
Focus: finalize safeguards; add names/titles handling; Unicode normalization; stricter back-translation outcome; ontology clarifications; delivery hygiene.
Project Brief additions

Names & Titles (NEW): preserve personal names/honorifics/titles; standard romanization (pinyin/IAST/Wylie); no invented or “upgraded” titles; lock one consistent form post first mention.

Existence/Non-existence policy (NEW §4.4)

Treat 有/無, 非有非無, 不落有無, 有無雙泯, etc., as technical ontological terms, not casual “have/has.”

Prescribed renderings: existence / non-existence, neither existence nor non-existence, not falling into existence or non-existence, both existence and non-existence are extinguished.

Preference guidance: 妙有 → “marvelous (wondrous) presence”; 真空妙有 → “true emptiness, marvelous presence.”

Contrastive gnomic lines covered, e.g.:
“有就是有，不是没有” → “Presence is presence, not absence.”

“有就是没有，没有就是有” → “Presence is absence; absence is presence.”

Step 0 polish & tougher criteria

0.1 Alignment & Ratio Sentinels: restated as “auto,” keep 3-in-a-row failure rule.

0.2 Structural Mirror & Quotation Parity: broadened marker parity to include dates, units, URLs.

0.3 Keyword Sweep: unchanged counts, but require explicit insertions for “Missing.”

0.4 Back-translation hardening: If drift/addition/omission remains unresolved ⇒ FAIL and re-run Step 0 (not just “note it”).

0.5 Read-Scan & Bilingual Audit: formalized as two sub-steps; any untraceable jump ⇒ FAIL.

0.6 Certificate Line fields expanded: now includes Quotes k/m (coverage of quoted/contrastive items) and clarifies each metric label.

Terminology & concordance

A.2 explicitly requires a mini-termbase + concordance sweep for consistency.

Technical delivery guardrails (NEW in Step B)

Unicode NFC normalization mandated for diacritics.

Ban “smart quotes”; preserve intended hyphen/dash distinctions; no hidden characters/HTML entities unless in source.

Formatting & delivery refinements

Reaffirm: no Markdown ordered lists for SegIDs; CSV fallback mandatory when tables appear.

Part break line standardized: “— End of Part X — [Ready for next part]”.

Staged clean-copy protocol reiterated (await “approve clean copy”; chunked delivery; safety stop on three consecutive >25% length deltas).

Guidance retained but clarified

Loanword-First Rule (IAST/Wylie with first-mention italics, roman thereafter) preserved and foregrounded.

Built-in Drift Tests kept (visions/no-marks; ālayavijñāna clause; trikāya completeness; powers sequencing; strict injunction diction).

Certificate Line (final form)

COMPLETENESS GATE: [PASS|FAIL] — Segments S=X/T=Y | Structures H/L/T/Q=H:a L:b T:c Q:d | Quotes k/m | Keywords found=k/missing=n | >25% deltas=p | Back-translation parity=[Confirmed|Drift fixed] | MQM(O/A/M/T)=o/a/m/t

Minor precision edits

Clarified that names/titles must not be “upgraded” in translation.

Explicitly added units/URLs to marker parity.

Kept entire doctrinal glossary lists unchanged; ensured 妙有 preference note aligns with §4.4.

Net Impact Summary
Completeness enforcement is now measurable and auditable (alignment + ratio + quote/contrast coverage + read-scan).

Omission risk is reduced via the quoted-string checklist, 3-in-a-row sentinel, and back-translation failure trigger.

Terminological integrity strengthened (mini-termbase + concordance).

Doctrinal precision improved for existence/non-existence and gnomic contrast lines.

Delivery hygiene improved (Unicode NFC, no smart quotes, CSV fallback, CommonMark safety).

Migration Notes (for reviewers adopting v5.4)

Build or enable simple alignment + ratio tooling; a spreadsheet suffices for small jobs.

Add a regex pass to collect quotes/contrast patterns (e.g., 「.?」|『.?』|“.*?” and (.+?)就是\1，不是(.+?)).

Prepare a lightweight termbase template (source term, canonical target, notes, first-mention gloss).

Normalize all drafts to NFC before QA to avoid diacritic drift across platforms.

Update: 20 October 2025:

I will now described the detailed change log for the three main prompts we have updated:
Prompt 1: The translation-only prompt (final version v5.3).

Prompt 3: The translation-with-commentary prompt (final version v4.1).

Prompt 6: The review and QA prompt (final version v6.0), which you've also referred to as Prompt 6.

Prompt 1: Translation-Only Workflow

This prompt evolved from a set of high-quality guidelines into a hardened, production-safe translation engine with non-negotiable checks for completeness and terminological fidelity.

v3.2 → v4.x Series

Focus: Introduce a professional workflow with modes, terminology management, and silent verification against authoritative sources.
Workflow Modes Introduced
MODE="prep_then_translate" and MODE="translate_only" were created to separate terminology harvesting from clean translation runs.

The PREP_PACK concept was introduced as a deliverable in prep_then_translate mode, allowing for the extraction and locking of a glossary.

Silent Pre-Flight Checks (NEW)
Mandated a "quiet" internal step to check uncertain terms against a specified list of authoritative resources (NTI Reader, CBETA, BDRC/BUDA, 84000). This grounded the translation in academic standards without cluttering the output.

Formalized Terminology
The detailed guidelines for translating key concepts (like "Disassociation," "Awareness vs. Mindfulness," and "self/Self") were carried over and preserved in full.

The extensive, unabridged terminology lists for Chinese and Tibetan were made mandatory.

Internal Self-Audit (NEW)
Introduced a silent MQM-style self-check (Accuracy, Terminology, Fluency, etc.) to be performed before output, improving the quality of the first draft.

v4.x Series → v5.3

Focus: Harden the prompt against omissions and structural errors by mandating 1:1 segment coverage, automated parity checks, and a strict, auditable workflow. This was the most significant architectural leap.
Configuration Block (NEW)
A formal CONFIGURATION block was added at the top to control core behaviors like NO_COMPRESSION: TRUE, ZH_VARIANT, and KEEP_QA_LINES.

Core Directives for Completeness (NEW & MANDATORY)
NO_COMPRESSION: TRUE: A hard flag forbidding any summarization of doctrinal content.

Mandatory Segmentation (SegID): The prompt is now required to parse the source into a SegID map (1...N) and translate 1:1. This is the primary defence against missing chunks.

Echo Counts: A mandatory pre-flight report (Detected X paragraphs / Y sentences) to confirm the source was fully analyzed.

Halt on Mismatch: A tripwire that stops generation and inserts a [MISSING] marker if a segment is dropped.

Verse/Quote Parity Must-Pass: An automated check and mandatory report (Verse parity: OK | Quote/URL parity: OK) that must pass before the final output.

Hardened Terminology & Doctrinal Rules
Tibetan Terminology Lock (NEW): Specific Chinese translations for key Dzogchen terms (rigpa → 明, lhun grub → 自然圆满（lhun grub）, gzhi → 基) were mandated.

Forbidden Variants Basket (NEW): Explicitly banned common but incorrect translations (e.g., forbidding 任运成就 for lhun grub).

Typography & Transliteration (Clarified): Formalized the use of IAST/Wylie with italics on first mention only, plus a strict adherence to Unicode NFC and no "smart quotes."

Doctrinal Guardrail Triggers (NEW): Added internal assertions for handling sensitive concepts like ālayavijñāna and visionary states (nimittas).

Workflow & Output Formalized
The output format was made stricter, dictating the exact sequence of the PREP_PACK, counts line, locked terms list, translation, and verification reports.

Term-Concordance mini-report (NEW): Added a mandatory report listing occurrence counts for locked terms to enforce consistency.

Net Impact Summary

Prompt 1 was transformed from an expert "translator" persona into a systematic, auditable translation engine. The introduction of mandatory segmentation (SegID), automated parity checks, and hard-coded failure states (like Halt on Mismatch) makes accidental omissions or terminological drift nearly impossible. It now operates with the process discipline expected in professional translation environments.

Prompt 3: Translation WITH Commentary

This prompt's evolution focused on integrating the industrial-strength chassis of Prompt 1 into the unique, multi-part structure required for scholarly commentary.

v3.2 → v4.1

Focus: A comprehensive merge, replacing the old workflow with the hardened framework from Prompt 1 (v5.3) while preserving the detailed, unabridged commentary and annotation structure.
Inherited Core Framework from Prompt 1
The entire CONFIGURATION block (NO_COMPRESSION, etc.) was ported over.

All CORE DIRECTIVES were integrated, including Mandatory Segmentation (SegID), Echo Counts, and the Verse/Quote Parity Must-Pass check.

The complete, unabridged MANDATORY GUIDELINES section was imported, including all case studies, detailed terminology lists, and advanced interpretive rules.

Merged Output Structure
The original Primary Output Requirement from v3.2 was meticulously merged with the new workflow. The prompt now requires the new automated reports (like the Parity Check Report and Term-Concordance Report) to be delivered within the existing structure of interleaved translation, detailed multi-part commentary, and the final scorecard.

The detailed breakdown of the Translator's Commentary (Introduction, Translation Choices, Contextual Explanations, etc.) was fully preserved and placed correctly within the new, more rigorous output sequence.

Modernized Workflow
The PRE-FLIGHT process (silent checks against NTI Reader, CBETA, etc.) was made a formal, silent step.

The concept of MODES (prep_then_translate vs. translate_only) was adapted for the commentary format.

Net Impact Summary

Prompt 3 was successfully upgraded to the same professional standard as Prompt 1. It now benefits from the same powerful defences against omissions and inconsistencies (SegID mapping, parity checks) while retaining its unique, rich output format for scholarly work. The merge ensures that both "translation-only" and "translation-with-commentary" tasks are performed with the same high degree of structural and terminological discipline.

Prompt 8: Translation Review & QA

This prompt evolved from a set of expert guidelines into a powerful, automated QA gateway with a non-negotiable "Completeness Gate" that programmatically verifies a translation's integrity before any qualitative review is allowed. The user sometimes referred to this as Prompt 6; we finalized it as Prompt 8.

v5.4 → v6.0

Focus: Transform the review process into a rigorous, automated QA system with a "Hard Gate" (Step 0) that programmatically enforces completeness, structural parity, and terminological consistency.
Step 0: The "Completeness Gate" (NEW & MANDATORY)
This entire step was created to serve as a non-negotiable prerequisite for the main review. The prompt is forbidden from proceeding until this gate is passed.

Alignment Table & CSV (NEW): Mandated as the first deliverable to ensure all segments are accounted for.

Segment Length Ratio Sentinel (NEW): An automated check that fails the gate if three consecutive segments have anomalous length ratios (e.g., outside 0.75–1.25), preventing silent summarization.

Quote/Marker Parity Check (NEW): An automated extraction and 1:1 mapping of all quoted strings and markers. Mismatches cause an automatic FAIL.

Verse Parity Check (NEW): A dedicated audit of line counts within blockquotes. Mismatches cause an automatic FAIL.

Term-Concordance Validator (NEW): An automated check against the glossary. Use of forbidden variants or inconsistent translations causes an automatic FAIL.

Hardened Reporting & Process Control
The Certificate Line (NEW): A mandatory, data-driven report that must be the first line of the entire output. It provides an at-a-glance summary of the QA checks. An imperfect certificate (FAIL, Forbidden Variants: >0, etc.) blocks the entire review.

Change-Log Discipline (NEW): For every FAIL cycle within Step 0, the prompt must produce a mini diff log explaining what was changed, where, and why, enforcing ISO-17100-style traceability.

Refined Guidelines & Audits
The unabridged content from v5.4 (Guiding Philosophy, Project Brief, Concept Policies, Reference Glossaries) was carried over in full.

SME Risk Register (Clarified): The requirement for [UNCLEAR – SME CHECK] was made more concrete: if any ambiguities are found, at least three specific items must be flagged for expert review.

Style & Typography Guard (NEW): Added an explicit checklist to the final report (Step B) to confirm adherence to IAST, Wylie, Unicode NFC, and no "smart quotes."

Net Impact Summary

Prompt 8 was transformed from an expert reviewer's checklist into an automated quality assurance system. The "Completeness Gate" mechanizes the most common sources of error (omissions, inconsistent terms, dropped quotes) and forces them to be fixed before any subjective stylistic review can begin. The mandatory Certificate Line and Change Log make the entire process transparent and auditable, aligning it with professional QA standards like ISO 17100 and MQM.

Update: 16 December 2025 — The "Ironclad" Anti-Substantialist Upgrade
Affected Prompts:
Prompt 1: v5.3 $\rightarrow$ v6.1

Prompt 2: v4.1 $\rightarrow$ v6.2

Prompt 6: v5.5 $\rightarrow$ v6.0

The Core Problem: The "D.T. Suzuki Effect"
We diagnosed a persistent issue where LLMs, despite having glossaries, would revert to outdated, reified translation choices (e.g., translating Ti 体 as "Substance" or Li 理 as "Noumenon"). This occurs because the model's training data is heavily weighted towards 19th/20th-century translations (like Suzuki or Legge) that use Kantian or Theosophical terminology inappropriate for accurate Buddhist phenomenology.

The Solution: Three-Layer "Ironclad" Defense
We introduced three new architectural mechanisms across all major prompts to override this training bias:
Dynamic Domain Switching: The AI now performs an initial logic check to determine if the text is Buddhist (Non-Essentialist/Anatta) or Essentialist (Hindu/Theist). It applies different translation rules based on this determination.

Pre-Flight Glossary Activation: Before translating a single sentence, the AI is now forced to scan the text and print out the specific terms it found and how it will translate them. This loads the rules into the active context window, preventing "instruction drift."

Negative Constraints (The "BANS"): We placed explicit, capitalized bans immediately before the source text input. The model is forbidden from using specific words (Substance, Noumenon, Soul, Entity, Shadows) unless the domain is strictly Theistic.

Specific Prompt Updates
Prompt 1: Translation-Only Workflow (v5.3 $\rightarrow$ v6.1)
Focus: Terminology enforcement and ontological safety.

New Feature: Added the Philosophical Guardrails section to distinguish between "Essence" (Empty Nature) and "Essence" (Soul).

New Feature: Added Step 1: Pre-Flight Check to the output order. The prompt fails if it does not list the detected terms first.

Hardened Glossary: Updated the JSON to strictly map Ti $\rightarrow$ Essence (with scope notes), Li $\rightarrow$ Principle, Miao You $\rightarrow$ Marvelous Presence.

Prompt 2: Translation WITH Commentary (v4.1 $\rightarrow$ v6.2)
Focus: Parity with Prompt 1 while preserving complex structure.

Integration: Ported the entire "Ironclad" chassis (Pre-flight, Anti-substantialism, Banned Words) into the commentary workflow.

Verification: Added specific checks in the "Self-Assessment Scorecard" to confirm that the translator did not slip into substantialist language (e.g., "Did I use 'Substance'? [YES/NO]").

Prompt 6: Translation Review & QA (v5.5 $\rightarrow$ v6.0)

Focus: Enforcing the new ontology during the review phase.

New Feature: The reviewer is now explicitly instructed to flag "Substantialist" terms (Substance, Noumenon, Entity) as Tier 1 Doctrinal Violations.

Refined Metrics: The scorecard now includes a specific "Substance Scan" to catch any reified terms that might have slipped through the translation phase.

Universal Context-Aware Mapping Protocol
Applicable Versions: Prompt 1 (v6.2), Prompt 2 (v6.4), Prompt 6 (v6.1)
New Feature: The "Master Rule" (Universal Logic)
We have introduced a Universal Context-Aware Mapping Protocol that applies a rigorous logic check to the entire text, not just a specific list of words.
The AI is now instructed to proactively detect any term—across the entire translation—where a common, Christian, or Secular definition conflicts with a Buddhist one. It must automatically enforce the Canonical Buddhist Register over Western/Theistic interpretations for all terminology.
Illustrative Examples of this Logic:
The following 8 rules are provided to the AI merely as non-exhaustive examples to demonstrate how it must apply this universal logic to every word in the text:
"Sin" vs. "Karma" (The Anti-Theistic Rule)
Logic: Buddhist causality is impersonal; "Sin" implies divine judgment.

Mapping: Strict Ban on "Sin." Map to "Transgression" or "Karmic Offense" (Zuì Yè 罪业).

"Rebirth" vs. "Resurrection" (The Samsaric Rule)
Logic: Buddhism posits cyclic existence ($samsara$), not a singular "born again" event.

Mapping: Strict Ban on "Resurrection" (Chóng Shēng 重生). Map to "Rebirth" (Lún Huí 轮回).

"Aversion" vs. "Disgust" (The Psychological Rule)
Logic: In Dharma, the "Second Poison" ($dvesha$) is ill-will, not just mundane distaste.

Mapping: Map to "Hatred/Ill-will" (Chēn Huì 瞋恚), avoiding secular "Disgust" (Yàn Wù 厌恶).

"Sense Restraint" vs. "Blocking" (The Yogic Rule)
Logic: Meditation ($pratyahara$) is the skillful gathering of attention, not a mechanical shutdown.

Mapping: Strict Ban on "Blocking/Closing." Map to "Gathering/Restraining" (Shōu Shè 收摄).

"Elements" vs. "Fluids" (The Physiological Rule)
Logic: Ancient Indian metaphysics ($dhatus$) refers to structural elements, not Western "humours."

Mapping: Strict Ban on "Body Fluids." Map to "Constituents/Elements" (Gòu Chéng Yào Sù 构成要素).

"Spirit" (The Essence Rule)
Logic: Spiritual essence ($non-dual awareness$) is distinct from secular "mental spirit" or "psyche."

Mapping: Strict Ban on Jīng Shén (精神). Map to "Spiritual/Numinous" (Líng Xìng 灵性).

"Contemplation" (The Practice Rule)
Logic: Zen enquiry ($huatou$) is distinct from Tantric visualization ($guanxiang$).

Mapping: Strict Ban on "Visualization" for Zen contexts. Map to "Investigate" (Cān Jiū 参究) or "Observe" (Guān 观).

"Void" (The Nihilism Rule)
Logic: Existential nihilism is distinct from "Empty Space" or "Open Sky."

Mapping: When referring to a dead blankness, map to "Nothingness/Void" (Xū Wú 虚无), avoiding positive "Empty Space" (Xū Kōng 虚空).