# SYS4 VM Mapping — Plan of Action

> **For the executing agent:** This is a reverse-engineering playbook. Work it phase-by-phase; each phase ends with a concrete, checkable deliverable. Verify claims against bytes before recording them. Validated seed data lives in `vm-map/`.

**Goal:** Decode the SYS4 bytecode into named instructions so the game logic can be re-implemented in Godot.

---

## ✅ BREAKTHROUGH (2026-07-05): the opcode set is already solved

**The prior "unpack AGE.EXE in Ghidra" critical path is no longer needed to disassemble scripts.** Kelebek1's decompiler ships a complete AGE opcode table that decodes this game directly.

**What was verified this session** (see `tools/validate_opcode_table.py`, run it to reproduce):

- Kelebek1/Eushully-Decompiler's `age-shared.cpp` contains an opcode table (`{op_code, label, argument_count}`) and a header parser that **explicitly handles the SYS4 signature** (`"SYS4"`, header length `0x3C`, cp932 XOR-0xFF strings) — this exact game's format.
- The instruction model: **code = a flat sequence of instructions; each instruction = `<opcode:u32>` followed by `argument_count` arguments, where every argument is a `<type:u32><value:u32>` pair. Instruction length in dwords = `1 + 2*argc`.** Inline strings sit *after* the code inside the `[0,F8)` region; stop decoding at the lowest string offset referenced (a type-2 arg, or op `0x64` arg 1).
- Applying that table to Himegari's scripts: **476 of 476 parseable scripts decode 100% clean — 1,463,788 instructions, 0 unknown opcodes, and all 37,392 inline-string arguments resolve to valid decoded strings.** (The 7 non-decoding `.BIN` are container-level non-scripts like `SYS4AB`/`SYS4INI`, different magic.)
- Himegari uses **248 distinct opcodes; 52 have semantic names** (in `vm-map/opcodes.toml`). The other 196 decode perfectly (known length) but have engine-internal names only (`u004xxxx`). **Caveat (measured 2026-07-06):** the named 52 are the dialogue/ADV core but cover only **72.6% of instruction volume**, not "the entire core" — the unnamed 27.4% is concentrated in the highest-frequency opcodes and must be partly addressed before Phase 4. See Phase 3's coverage correction.

**This resolves the header unknowns too.** Kelebek's `BinaryHeader` struct maps my F0–F12 exactly: `F0`=local_integer_1, `F1`=local_floats, `F2`=local_strings_1, `F3`=local_integer_2, `F4`=unknown, `F5`=local_strings_2, `F6`=sub_header_length(0x1C), then the three (length, offset) table pairs. The "flag fields" were **local-variable counts**. Arg `type` codes: 0=immediate, 1=float, 2=string, 3=global-int, 4=global-float, 5=global-string, 6=global-ptr, 8=global-string-ptr, 9=local-int, A=local-float, B=local-string, C=local-ptr, D=local-float-ptr, E=local-string-ptr.

**Consequence:** Unpacking `AGE.EXE` (still packed — see appendix) drops from *the blocker* to an *optional enrichment* used only to name the 196 unnamed opcodes' fine semantics, and even that has a cheaper dynamic alternative.

**Provenance / sources in `vm-map/`:** `kelebek1-age-shared.cpp` (the opcode table), `kelebek1-disassembler.cpp` (the parser), `opcodes.toml` (validated table filtered to what this game uses), `opcode-leads.json` + `small-script-listings.md` (this session's static analysis, now confirmed).

---

## Global constraints

- **Python:** `py -3.11 -X utf8 …` always (Shift-JIS output needs utf8 mode on Windows).
- **Authoritative copies:** 49 loose game-folder script `.BIN` files shadow `extracted/DATA1/` copies at runtime; two engine BINs are root-only. Target the game-folder copy where both exist. `sys4load.load()` is copy-agnostic; `paths.scripts()` resolves the override.
- **Units:** all script offsets/counts are DWORDS (×4 bytes), relative to body start `0x3C`.
- **Instruction rule:** `len_dwords = 1 + 2*argc`; args are `(type,value)`; **stop code decode at the first inline-string/array offset**, not blindly at `F8`.
- Record confidence per finding (confirmed-by-bytes / confirmed-by-runtime / hypothesis).

---

## Phase 1 — Port the opcode table into `sys4load.py`, disassemble everything *(✅ DONE 2026-07-05)*

**Deliverable:** `sys4load.py` emits real named instructions; every script disassembles with zero unknown opcodes. **Achieved: 481/481 DATA1 scripts decode fully clean; MENU.BIN and SC0030.BIN verified by hand.**

- [x] **1.1 — Embed the opcode table.** Full Kelebek table (548 entries) transcribed to `tools/age_opcodes.py` (`OPCODES`, `ARG_TYPES`, `CONTROL_FLOW`, `is_label_argument`), generated from `vm-map/kelebek1-age-shared.cpp`.
- [x] **1.2 — Replace the T3-chunking stub.** `sys4load.py` now has `decode_code()` (the `1+2*argc` walker with shrinking `code_end`) and a rewritten `render_listing()` that prints mnemonics, typed operands, inline strings, and `label_xxxx:` control-flow anchors.
- [x] **1.3 — Validate.** `tools/sys4load.py ../../extracted/DATA1 --validate` → **481/481 parsed clean, 481/481 opcode-decode clean.** MENU.BIN: 148 instrs, `set-font "ＭＳ 明朝"` + `comment` strings correct. SC0030.BIN: 11,951 instrs, `show-text` shows dialogue inline. (`tools/validate_opcode_table.py` still reproduces the standalone 476/476 over the merged root+DATA1 set.)
- [x] **1.4 — Regression-guard:** container `--validate` still reports 481 clean, 0 failures, 0 impure tags.
- [x] **1.5 — Disassembler is the artifact.** `sys4load.py <file>` prints the full listing; `--json` now includes decoded `code` (with `--json` + `to_dict(with_code=True)`), instruction counts, and decode-clean flag.

**Note:** 7 root `.BIN` are non-script engine indices with different magic (`SYS4INI` = `S4IC422`, `SYS4AB` = `S4AB`, etc.) — correctly rejected by the container parser, not scripts.

## Phase 2 — Extract data tables + dialogue *(✅ mostly DONE 2026-07-06)*

**Deliverable:** game database as JSON + full translatable dialogue corpus. Structure spec: `docs/PROJECT-STRUCTURE.md`. Extractors: `tools/extract_phase2.py`, `tools/extract_init.py`.

- [x] **2.0 — Project structure.** Established `docs/`, `build/{disasm,text,data,scripts-json}/`, `godot/`; game install stays read-only in place. Also relaxed the loader magic check to the `SYS4` family (`SYS4424` patch scripts now parse — was silently skipping 5 scripts).
- [x] **2.1 — Text corpora.** `tools/extract_phase2.py` → 481/481 scripts: full disassembly (`build/disasm/*.asm`), per-script strings, `build/text/dialogue.jsonl` (**30,057 show-text lines** — the translation corpus), `build/text/strings.jsonl` (38,449 strings tagged by source opcode), `build/manifest.json`.
- [x] **2.2 — `*INIT` data and rule sources → JSON.** `tools/extract_init.py` auto-detects shape (`name`/`numeric`/`footer`/`mixed`/`rules`/`dispatch`/`banked`) → **SKINIT (131 skills), ITINIT (287 items), EBINIT (277 units), OBINIT (46 object definitions), CIINIT (24 character-information profiles)** [name: sparse one-based name/description/fields, with CIINIT's dedicated profile registry], **CGINIT (851 sparse gallery images)** [numeric: full image, optional save/stage preview, thumbnail sheet/slot, and variant ordinal in a reserved 2,000-row layout], **ALINIT (107 sparse alchemy recipes)** [numeric: output, level/story gates, point cost, and four paired ingredient slots in a reserved 1,000-row layout], **AFINIT (13 authored affinity rows plus tuning/facility curves)** [name/footer specialization], **CTINIT (five 70-cell name-entry pages)** [string-matrix specialization], **CVINIT (13 character-voice settings)** [numeric: preview assets, slot-to-unit joins, and inverse suppression-setting map], **LAINIT (20 shipped terrain definitions)** [name: sparse strings, topology/rendering, combat-stat, traversal-skill, and texture-fallback joins], **MPINIT (1,472 sparse terrain rows joined to 66 stage maps)** [footer: 53-column doubled-coordinate atlas], **SPINIT (eight 15-slot H-scene gallery pages)** [numeric: INIT2 thumbnail-sheet and SYS4INI scene joins], **TRINIT (21 training actions)** [name/string-matrix: eligibility, cost, stat/alignment/progress effects, rewards, and ten event slots], **STINIT (74 stages)** [mixed: selector-dispatched strings/scalars/buffer cells/footer arrays], **CCINIT (71 class-change rules over 33 units)** [rules: unit/level/state predicates plus title/cost/stat/skill effects], **SCINIT (1,209 final decision rows)** [dispatch: scene resource plus authored chapter metadata and overwrite history], and **RTINIT (172 routine sets)** [banked: twenty movement/battle step banks, provider joins, and overwrite history]. Validated; see `build/data/README.md`. Raw addresses remain bytecode provenance; confirmed semantics come from `vm-map/globals.toml`.
- [x] **2.3 — Extract `STINIT`'s 74 sparse stage records.** The mixed mode identifies the dominant `scjump_progress_a` dispatch, recovers 29 preallocated buffer layouts (including six consumer-confirmed row strides), and keeps four condition strings, six scalars, fixed-buffer writes, and all 1,396 footer-array copies separated by stage id. Profiles supply population/value and direct-consumer evidence. Header/map semantics cover conditions, BGM, turn/replay/clear-reward settings, and map texture overrides. The object join assembles 2,312 placements with type/position/gates, OBINIT names and available descriptions, 604 reinforcement schedules, and typed initial-faction, teleport, treasure, card-list, non-triggering-faction, and initial-object-state payloads; FIELD's dedicated special-spawn path proves the final three populated type-27 tagged writes are engine-dead, so they remain visible as ignored provenance rather than unresolved semantics. The 30-cell enemy family contributes 1,378 joined spawns with unit/faction, placement, difficulty/story gates, level scaling, weighted selection, difficulty-specific movement/battle routine sets, and 485 first-clear-only gates. Raw address views remain alongside generated `semantic_fields`.
- [x] **2.3a — Extract `CCINIT`'s 71 class-change rules.** Rules mode preserves source-order guards and detects the unit id, minimum level, clear applied-state slot, selected title/level, deployment-cost delta, named fourteen-stat bonuses, awarded SKINIT skills, and state slot set by each rule. EBINIT/SKINIT definition joins and the common global semantic projection coexist with raw addresses. CALCCC/ADDEXP establish the apply/report ABI; EVOLVE explains the two level-independent empty-title Lily rules. The generated profile covers 33 units, 69 titled rules, 30 skill awards, three used promotion slots, and 19 populated output fields.
- [x] **2.3b — Extract `SCINIT`'s scene-dispatch registry.** Dispatch mode recognizes 2,179 alternating writes to two 10,000-cell arrays, preserves all 710 overwritten decision ids, and emits 1,209 final rows. The primary column maps decisions to 135 packed SYS4INI ids, all resolved to numbered SC scripts; the parallel column is authored chapter metadata. All 847 live SCJUMP decisions join to rows and 844 final chapter tags agree with the independently decoded paths; three legacy/stale mismatches remain explicit.
- [x] **2.3c — Extract `RTINIT`'s routine-step banks.** Banked mode recognizes twenty parallel 1000-by-20 tables and emits 172 sparse routine-set rows with all 3,336 source assignments and 3,307 final cells. Movement banks join all 1,043 steps to all 19 used RTN_M providers (22 dispatchable); battle banks join fourteen steps to all four RTN_B providers. Activation percentages, progress gates, and required/forbidden story flags are consumer-proven; six empty banks remain explicit. RTN_M001/002/003/004/005/006/007/008/009/010/011/012/013/014/015/017/051/052/061 add selector-specific progress, randomized roaming, object-slot, coordinate, enemy/ally, treasure and object search, Healing Feather, waypoint, faction-terrain, retreat, normal-attack routing, offensive target/action selection, and immediate allied-healing semantics. All 977 populated movement-parameter cells are classified: 974 semantic inputs and three M001/M008 cells proven unread; thirteen unwritten defaults are projected separately while the generic raw banks remain intact.
- [x] **2.3d — Extract `ALINIT`'s alchemy recipes.** Numeric specialization recognizes seven recipe-indexed structures and emits 107 sparse recipes from all 914 static writes. ALCHEMY proves the output, minimum level, required/forbidden story gates, point cost, and four paired ingredient-id/quantity slots. Every output and all 286 ingredient references join to ITINIT; raw coordinates remain beside the semantic and nested ingredient views.
- [x] **2.3e — Extract `AFINIT` affinity/progression data and `CTINIT` name-entry characters.** AFINIT classifies 27 element strings and 54 footer arrays into thirteen signed effectiveness rows, paired tuning bonus/cost curves, and three facility-progress rows. CTINIT preserves 273 authored characters and 77 empty cells across INPUTNAME's five reserved 70-cell pages. Both schemas account for every instruction and retain raw coordinates.
- [x] **2.3f — Extract `CVINIT` character-voice settings.** Numeric specialization accounts for all 37 static writes as thirteen preview OGG assets, twelve CONFIG-setting-slot-to-EBINIT-unit joins, and the exact inverse unit-to-suppression-setting map. CONFIG proves preview, row-label, persisted speaker-seen unlock, and setting-edit behavior; the shared CNINIT/CVINIT voice chain proves the inverse map.
- [x] **2.3g — Extract `MPINIT`'s stage-terrain atlas.** Footer specialization classifies all 1,472 row copies as columns 1..50 of one stride-53 sparse atlas, with destination-derived grid Y=2..1600 and 127 implicit-zero rows. FIELD proves that STINIT2's tile bounds are doubled before copying into the mutable current-stage grid. The schema joins 66 stage definitions to 53 unique rectangles, preserves eight shared-map groups and 47 border-context cells, and resolves terrain ids through LAINIT's names, texture slots, area-fill flags, and layout classes.
- [x] **2.3h — Extract `LAINIT`'s terrain definitions.** Name specialization classifies all 98 instructions into twenty shipped terrain ids inside a reserved thirty-row registry: seventeen names, five effect descriptions, four parallel terrain arrays, a ten-column signed combat-stat matrix, five SKINIT traversal/reveal requirements, and ten SYS4INI-resolved shared texture fallbacks. CALCBTPARAM, MVSEEK, FIELD, INFOAF, and DRAWMAP prove the complete consumer contract.
- [x] **2.3i — Extract `SPINIT`'s H-scene gallery registry.** Numeric specialization classifies all 118 writes as eight fifteen-slot HMODE pages with two implicit trailing zero cells. Every row joins to INIT2's `SO027A` through `SO027H` thumbnail sheet and every populated cell resolves to an `SP*.BIN` resource. HMODE proves the opcode-0x19d availability-filter and indirect-call contract.
- [x] **2.3j — Extract `TRINIT`'s training-action registry.** Name/string-matrix specialization classifies all 365 instructions into 21 six-text-slot actions and nineteen contiguous numeric families. TRAIN proves story/level/alignment/progress/stat/item/skill gates, spirit and fourteen-stat effects, fractional alignment/progress changes, rewards, execution counts, and event dispatch. ITINIT/SKINIT resolve every gate and reward; all 75 event cells resolve through SCINIT, while GAMESTART proves their restored story-flag role.
- [x] **2.4 — Partial global-var map BUILT + wired into the disassembler.** `tools/global_map.py` → `build/global-var-map.{json,md}` (16,354/49,435 globals labelled: string tables, `*INIT` field arrays, 122 record tables w/ strides, current-entity index pointers). `sys4load` renders the labels inline (`=rec[s30]`, `=current-entity-index?`). See `docs/name-resolution.md`.
- [ ] **2.5 — Grow the global-var map (future, incremental).** Static first: RTINIT is closed at 1,043/1,043 movement steps; the RECOVER/ILINIT condition ABI, ALINIT recipes, AFINIT affinity/progression tables, CTINIT name palette, CVINIT character-voice registry, LAINIT terrain definitions, MPINIT stage-terrain atlas, SPINIT H-scene gallery, and TRINIT training actions are closed; ITMES/SKMES/VIMES/EIMES/CIMES/MAMES are joined to their definitions; and the non-table INFOMES/MES ABIs are classified. Next, audit CDINIT's card-generation registry and its STINIT/FIELD consumers, then use Frida only for semantics static consumers cannot settle. Full detail: `docs/name-resolution.md` → "Future step — growing the map". Packed `call-script` ids, SCJUMP decision-to-scene dispatch, the shipped RTINIT movement-provider join, and the six completed message joins are resolved.

## Phase 3 — Name the unnamed opcodes *(top ~20 BEFORE Phase 4; the rest on demand)*

> **⚠️ Coverage correction (measured 2026-07-06).** The earlier framing — "52 named ops
> cover the entire core, name the other 196 lazily" — is **overstated**. Across the full
> corpus (1,503,166 instructions, all 481 scripts), **named opcodes are only 72.6% of
> instructions; the 195 unnamed `u004xxxx` ops are 27.4%** — and that 27% is front-loaded
> into the *most common* opcodes, not a deferrable long tail. The top unnamed ops by
> frequency: `0x1f4`/`0x1f5` (**60,297 each** — equal counts → a begin/end or push/pop
> pair, both zero-arg), `0x1d5` (34k), `0x1bc` (27k), `0x71` (26,445 — *exactly* the
> corpus T1 label-table entry count, so it's the **label-definition pseudo-op**, nameable
> by structure for free), `0x1a2` (18k), `0x7a` (17k, argc 3, follows arithmetic →
> computational), `0x1d2` (17k). **A Godot VM hits these in the first few instructions of
> any script.** So naming the top ~20 is a *prerequisite* for Phase 4, not a lazy
> follow-on. Only the genuine long tail (rare ops) is deferrable. Reproduce the measurement
> by iterating `sys4load.load` over the corpus and bucketing `ins.opcode` against
> `age_opcodes.OPCODES` (label starting `u00`/`dev_ukn` = unnamed).

> **⚠️ Named labels are from a *different* AGE title.** The 52 semantic labels are
> transcribed from Kelebek's table for a *later* AGE game. The opcode **number + argc** are
> validated for Himegari (481/481 clean decode proves structure), but the **semantics are
> not independently verified**. The ADV/text core is empirically safe — the 30,057-line
> `build/text/dialogue.jsonl` is proof that `show-text`/`end-text-line`/the string
> mechanism are right, and arithmetic/control-flow labels are corroborated by operand-type
> and jump-target consistency. The exposure is the **effectful named ops you can't see in
> text output** (`play-voice 0xc4`, `draw-texture 0x1fb`, sound/UI/draw ops) — Frida-confirm
> those against Himegari before the VM relies on them; don't assume them.

**Do this before Phase 4:** name/classify the ~20 highest-frequency unnamed opcodes.
Most fall to free inference (3.0); a few opaque effectful ones want a Frida session;
computational ones suit Unicorn. Everything below still applies — it's the *ordering* that
changes, not the toolkit. The genuine rare tail stays lazy (name on demand).

- [x] **3.0 — Inference pass DONE (2026-07-06).** Classified the top 21 unnamed opcodes →
  **instruction coverage 72.62% (named) → 96.94% (classified)**; ~90.5% is VM-handleable by
  inference alone. Tooling: `tools/opcode_context.py` (evidence gatherer). Results:
  `vm-map/opcodes.toml` (per-op evidence + provenance), `tools/age_opcodes_himegari.py`
  (`INFERRED` dict consumed by the disassembler + future VM), `build/opcode-coverage.md`
  (tiers + Frida/Unicorn shortlist). `sys4load` now renders inferred names (verified: MENU's
  `label-def 0x71` land exactly on its T1 targets). Key findings: `0x1f4`/`0x1f5` = stmt
  begin/end brackets, `0x1d5`/`0x1bc`/`0x1bf` = block markers (all zero-arg no-ops); `0x71`
  = label-def (count == T1 size); `0x21b`/`0x1d2`/`0x258` = tentative-no-op statement metadata
  (harness-verify); `0x7a` = ADV text param, `0x202/0x203/0x1f7/0x1fa/0x217/0x218/0x21a/0x1ff`
  = draw/UI, `0xb6` = audio, `0x215` = count/search — the effectful/computational Frida/Unicorn
  shortlist. Reserve live tools for those; rare tail (3%) stays lazy.

### 3.1 — Frida: dynamic observation *(primary tool for effectful opcodes)*

Frida injects a JS engine into the **running** game and hooks functions live. It sidesteps the packer (memory is already decrypted by the time you attach), gives ground-truth behavior, and lets you correlate an opcode with its on-screen/audible effect — the only reliable way to name rendering/audio/input/save/UI handlers. Two stages:

- [ ] **3.1a — Locate the dispatch loop.** Kelebek's `u004xxxx` addresses are from a *different* AGE title and will NOT match Himegari's `AGE.EXE`, so find Himegari's dispatch first. Best anchor: search process memory for a known script's opening opcode sequence (you have every script decoded), set a **hardware read breakpoint / `MemoryAccessMonitor` guard page** on its first opcode dword; when the VM fetches it, the instruction pointer is inside the dispatch fetch. Alternate anchors: breakpoint a winmm/DirectSound call and trigger `play-voice` (0xC4), then walk the stack back; or pattern-scan for the bounds-check + `call [table + opcode*4]`. **Payoff:** read the jump-table base → you get the handler address for all 548 opcodes in Himegari at once.
- [ ] **3.1b — Instrument + correlate.** `Interceptor.attach` the dispatch (or a specific handler); log opcode + operand `(type,value)` pairs (read from the bytecode pointer — layout known) + effect. Three correlation techniques: **API** (hook a basket of D3D9/winmm/user32/file APIs; see which an unknown handler calls), **behavioral** (trigger one in-game action, diff the opcode trace vs. baseline to attribute ops to subsystems), **memory** (log which global/local var-bank slots — sized by header F0–F5 — the handler reads/writes).
- **Setup / gotchas:** `pip install frida-tools`; **attach to the already-running game** (`frida AGE.EXE`) after the title screen rather than spawning — this skips the packer's startup anti-debug. 32-bit x86 target. Japanese locale required to run. Eushully's protector *may* detect Frida's injected thread; if it trips, quiet it (ScyllaHide-style hooks or `frida-gadget`).

### 3.2 — Unicorn: microexecution *(complement for computational opcodes)*

Unicorn is a bare CPU emulator (no OS). It is the **better** tool for the *pure-computation* handlers — arithmetic/bit/string/array helpers and especially the `CALC*` damage/stat formulas — where you want the *exact* operation, not a label. It is **blind** to effectful handlers: the instant one calls D3D9/winmm/file APIs it runs into unmapped code and stubbing tells you nothing (the effect *is* the meaning). Do not use it as a Frida replacement.

- [ ] **3.2a — Microexecute a handler.** Map the handler's code + a synthetic VM state (variable bank + operand), run from entry to `ret`, read back what changed; sweep inputs to recover the formula deterministically, offline.
- [ ] **3.2b — Preferred combo: Frida-snapshot → Unicorn-replay.** Use Frida (3.1a) to find handler addresses and dump the relevant memory (code + var banks + globals) at a known-good moment (e.g. mid-battle); load that snapshot into Unicorn and microexecute individual handlers with input sweeps. Gets Frida's context-setup for free + Unicorn's determinism. **Caveat:** microexecution only recovers behavior that's a pure function of the captured state — if a handler reads a global you didn't snapshot, results are wrong silently. Fine for pure ops; a rabbit hole for stateful ones (leave those to live Frida). Needs a decrypted image to feed (a dump, or bytes pulled via Frida) since `AGE.EXE` is packed.

### 3.3 — Cross-reference siblings *(free, do alongside 3.0)*
- [ ] Kelebek's labels come from a later AGE title; marcussacana/EushullyEditor targets *Kamidori* (same SYS4 era). Diff their handler notes for the specific opcodes you need.

### 3.4 — Static unpack + Ghidra *(last resort)*
- [ ] Only if the above stall. See appendix — dump the decrypted image, load in Ghidra, read the handler at its address. High effort; reserve for genuinely opaque ops that Frida/Unicorn can't pin down.

## Phase 4 — Godot re-implementation

**Deliverable:** the AGE VM running Himegari scripts in Godot.

> **Prerequisite:** don't start Phase 4 against a blank opcode set — the top ~20 unnamed ops
> (Phase 3 preamble) are hit in the first few instructions of any script. Do that thin
> naming slice first, or bring-up stalls immediately on `0x1f4`/`0x71`/etc.

- [ ] **4.0 — Stand up the validation harness *first* (before writing VM opcodes).** The
  strongest correctness oracle already exists in `build/`: for a given scene script, the VM's
  emitted `show-text` sequence must match that file's lines in `build/text/dialogue.jsonl`.
  Wire this as an automated diff (drive one `SC####` script → collect show-text → compare to
  the 30k-line corpus filtered by `file`). This turns "is the VM right?" into a per-scene
  regression test and catches control-flow/branch bugs (wrong jcc → wrong dialogue order)
  early. Extend later to assert extracted-table reads (SKINIT/ITINIT/EBINIT JSON) once
  data-driven opcodes come online.
- [ ] **4.1 — Re-implement the VM** (GDScript/C#): a dword-fetch loop, the core opcodes (arithmetic, comparisons, `jmp`/`call`/`jcc`, `mov`, string ops) **plus the high-frequency unnamed ops named in Phase 3**, the global/local variable banks (sized by the header's F0–F5 counts), and the ADV layer (`show-text`/`end-text-line`/`wait-for-input`/`set-font`/`play-voice`/`draw-*`). Note the named set is only ~73% of instruction volume — budget for the unnamed remainder. The bytecode-heavy design (damage calc, dungeon loop, battle flow are all scripts) makes re-implementation the right call over transpilation.
- [ ] **4.2 — Fill opcodes on demand** from Phase 3's genuine long tail as scripts exercise them (the top ~20 are already done as a Phase 3/4 prerequisite).
- [ ] **4.3 — Deferred:** save-file format (reverse `SAVE.BIN` only if the port must read existing saves).

---

## Appendix — `AGE.EXE` is packed (relevant to Phase 3.4, and as the image source for 3.2)

Verified this session (`tools/pack_check.py`): 32-bit PE, code sections at max entropy (8.00), blank section names, IAT RVA 0, no plaintext anchors. **`SYS4AB.BIN` (magic `S4AB`) is a dead end for static analysis — it decrypts to `AGE.EXE` byte-for-byte** (2026-07-06): 0x2c-byte header (`"S4AB"` + version + `0x0010E000` size dword ×3 + an 8-byte key/hash field), then the payload is a trivial **XOR-`0xFF`** of the *same packed* `AGE.EXE` (`bytes(x^0xFF for x in payload) == AGE.EXE`, exact). So it is **not** a patched/unpacked VM — both on-disk engine images are the identical packed binary, and VA `0x421160` (any handler) is entropy-8.00 garbage in both. The real handler code exists **only unpacked in the runtime heap** (`~30 MB r-x @ 0x62411000`, nonstable base per run — see `docs/global-memory-re.md`). Static Unicorn/Ghidra therefore requires a **runtime dump** of that region, or hook it live.

Static analysis therefore requires a **runtime dump first** — you're dumping for *analysis* not redistribution, so don't chase OEP: launch to the title screen (Japanese locale required), then dump the decrypted image and load it in Ghidra. Tools: **PE-sieve** (CLI, agent-drivable: `pe-sieve.exe /pid <PID> /imp 3`) or **x32dbg + Scylla + ScyllaHide** (GUI, handles the anti-debug). Validate the dump by confirming `SYS4422`/`.BIN`/`DATA1` now appear in plaintext. **But prefer Frida dynamic hooking (Phase 3.1) — it avoids the unpack entirely.**