WP-006: Phantom archive format verification (Gen 1 vs Gen 2)

Status: 📦 Queued Phase: 1 — Format Research (follow-on to the Engine Lineage finding in WP-005) Depends on: — Companion EC: — (single hex dump; no checklist needed) Estimated effort: 1 minute

---

Goal

Determine whether Scooby-Doo! Phantom of the Knight’s archive format is Gen 1 (TGIFILE.ART, same as Showdown) or Gen 2 (MMFW wrapper, same as Jinx). The Rich Header data classifies Phantom as Gen 1 by toolchain (identical to Showdown’s Linker 5.10 build 8047), but the archive format is the other half of the generation classification and hasn’t been inspected on the Phantom disc yet.

Background

The Phase 1 Engine Lineage finding identified three format generations: Gen 1 (Showdown / Phantom predicted), Gen 2 (Jinx), Gen 3 (Case File #1

• Case File #2 predicted). Toolchain build numbers and archive format coincide across generation boundaries — same Linker 5.10 build 8047 appears with TGIFILE.ART (Showdown); same build 8168 appears with MMFW (Jinx); both 8168 and 8797 appear with MMFW (Case File #1).

Phantom’s Rich Header matches Showdown’s exactly (build 8047). The prediction is that Phantom’s archives are therefore also Gen 1 TGIFILE.ART. But Phantom shipped in Aug 2001 and Jinx in Oct 2001 — TerraGlyph cut the Gen 2 format somewhere in that window, and we don’t know yet which side of the cut Phantom lands on.

Locking this in matters for:

• docs/01-VISION.md Engine Lineage table — currently shows Phantom as Gen 1 with archive format flagged unverified
• docs/formats/tgifile-art.md Cross-title verification section
• The Phantom row in the eventual ADGameDescription detection table’s generation flag

Scope

In scope:

• Mount Phantom’s ISO (or copy the archive files to tools/exes/phantom/ per the project convention)
• Identify archive file(s) at the disc root or under scooby\
• Run Format-Hex on the first 16 bytes of the largest archive file
• Compare against:
  • Gen 1 marker: 45 00 00 00 (Showdown’s TGIFILE.ART opening — group_count = 69)
  • Gen 2 marker: ASCII MMFW at offset 0
• If the first-marker check is Gen 1, run two additional structural checks (per WP-003 baseline — the prior version of this WP would have given a Gen 1 verdict on coincidence-prone evidence). These turn a 4-byte coincidence into a structural verdict:
  1. Header layout check — Phantom’s TGIFILE.ART should match the corrected Gen 1 layout from tgifile-art.md §Header layout: asset_count at offset 0x0118 (uint32LE, plausible small integer in the low-thousands range; Showdown’s is 453). If it’s at 0x011C instead, Phantom is a Gen 1.5 variant worth recording, not vanilla Gen 1.
  2. Name-table check — bytes at offset 0x0F60 should begin with the 68-byte name-record pattern (4-byte uint32LE id with type-tag high byte in {0x10, 0x20, 0x31–0x3F}, followed by an ASCII string starting with ROOM_, OBJ_, or ANIM_). If the first record at 0x0F60 decodes as a ROOM_* name, Phantom is structurally Gen 1.
• Update the two doc references with the verdict (plus any structural surprises)

Out of scope:

• Decoding any Phantom archive payload (that’s Phase 2+ once Showdown’s TGIFILE.ART payload format is cracked)
• Bit-level comparison of Phantom’s Scooby.exe against Showdown’s (interesting trivia, not needed for this verdict)
• Anything else on the Phantom disc

Dependencies

• Phantom ISO mountable (already on disk at C:\pcloud\SCOOBY\Scooby Doo Phantom of the Knight.iso)
• Format-Hex (built into PowerShell)

Exit criteria

1. The verdict is binary — Phantom is Gen 1 or Phantom is Gen 2 — recorded in both:
   • docs/01-VISION.md Engine Lineage table (the Phantom row’s “Archive” cell stops saying “unverified”)
   • docs/formats/tgifile-art.md Cross-title verification section (the Phantom bullet stops saying “predicted”)
2. The hex bytes that produced the verdict are pasted into the appropriate format spec (tgifile-art.md if Gen 1, mmfw-container.md if Gen 2).
3. If verdict is Gen 1: the two structural checks from §Scope (header layout per WP-003 corrected offsets + name-table sanity at 0x0F60) both pass, OR any deviation is documented as a “Phantom Gen 1.x variant” finding in tgifile-art.md.

Deliverables

• Updated docs/01-VISION.md — Engine Lineage table Phantom row
• Updated docs/formats/tgifile-art.md — Cross-title verification
• Possibly updated docs/formats/mmfw-container.md — only if Phantom turns out to be Gen 2

Notes

• Expected result: Gen 1. The build-pipeline upgrade that introduced Gen 2 (newer linker 8168, dropped MASM 6.14 + cvtres) coincides with the archive format change, and Phantom’s Rich Header shows the old toolchain (linker 8047). If the format follows the toolchain, Phantom is Gen 1.
• If the result is unexpectedly Gen 2, that’s meaningful: it would mean the engine evolved during a single linker build cycle, with archive format changing at a different cadence than toolchain. Worth a one-paragraph follow-up in docs/formats/scooby-exe.md Findings → cross-generation analysis.
• Worth caching while you’re at it: since you’ll have the Phantom disc mounted, copy Scooby.exe (and the archive file you Format-Hex’d) into tools/exes/phantom/ per the tools/exes/ README. Future analysis won’t need to re-mount.