WBML-0002 — addon state stays isolated between world saves

✅ Verified 🧪 WorldBox Modding Lab WBML-0002

WBML-0002 asked a different question from WBML-0001:

If two world saves use the same addon and the same logical data keys, does one world’s addon state appear in the other world?

Why this needed a separate probe

WBML-0001 verified:

same save + full process restart → values return

That did not prove that the values were isolated from other worlds. A bad implementation could still have appeared persistent while actually leaking through static/runtime state.

WBML-0002 therefore used two different saves and two different current-run signatures.

Harness correction before the valid run

The first 0.0.2 harness had a methodological bug: a failed Step C did not prevent Step D from printing a final success line.

That run was rejected.

0.0.2-fix1 added:

strict A → B → C → D state machine
unique run token
step flags
final gate requiring A=PASS B=PASS C=PASS D=PASS
failure lockout

This matters because a probe is only evidence if its own verdict logic is trustworthy.

Verified environment

WorldBox:          0.51.2
build:             719
git:               build-719@5dec
NeoModLoader:      1.2.0.1
PoliticalWorldAPI: 1.14.0
Lab:               0.0.2-fix1

Current-run token:

af599151c6da

Sequence

World A
→ write A signature
→ save
→ load World B
→ confirm A signature absent
→ write B signature
→ save
→ load World A
→ confirm B signature absent + A values return
→ load World B
→ confirm A signature absent + B values return

The probe tested:

marker
int
Unicode string
bool
float
addon-private kingdom tag
shared kingdom tag

Step results

A — write World A

A-IMMEDIATE RESULT: 6/6 DATA PASS | marker=PASS => PASS
STEP A RESULT: PASS

B — verify clean World B and write B

Before writing B, the probe scanned 21 kingdoms:

WORLD B CLEAN CHECK: PASS
no current-run WORLD A signature found across 21 kingdoms

Then:

B-IMMEDIATE RESULT: 6/6 DATA PASS | marker=PASS => PASS
STEP B RESULT: PASS

C — return to World A

RETURN A: current-run WORLD B signature absent => PASS
A-RETURN RESULT: 6/6 DATA PASS | marker=PASS => PASS
STEP C RESULT: PASS

D — return to World B

RETURN B: current-run WORLD A signature absent => PASS
B-RETURN RESULT: 6/6 DATA PASS | marker=PASS => PASS
STEP D RESULT: PASS

The corrected final gate then required every step:

FINAL GATE: A=PASS B=PASS C=PASS D=PASS
FINAL RESULT: WORLD ISOLATION VERIFIED FOR THIS RUN.

Verified claim

For the tested environment, PoliticalWorldAPI kingdom state used by the probe stayed isolated between the two tested world saves during the same WorldBox process.

The current-run A signature did not appear in B, the current-run B signature did not appear in A, and each world’s own values returned when that save was loaded again.

Evidence boundary

This result does not by itself prove:

party-private data isolation;
the exact physical save file/database used by each value;
isolation for every possible PoliticalWorldAPI data type;
behavior across future WorldBox/NML/Political World versions;
a full-process-restart isolation sequence.

Full-process persistence is a separate verified result from WBML-0001.

Methodology lesson

A failed probe harness is still valuable research material if the bad verdict is rejected rather than promoted.

The corrected rule is simple:

final result = conjunction of every required step

not:

last step passed → whole experiment passed

This rule now belongs to the Lab’s evidence methodology, not just this one test.