WBML 0.3.0 — Capability Verification Atlas
Final status: CLOSED PASS
Canonical probe: 0.3.0-fix1
Run: 82c23a7c8916
Final matrix: 28 PASS / 0 FAIL / 0 SKIP, P–F PASS.
Research question
Section titled “Research question”From the 0.2 structural surface, which zero-argument non-void methods can be structurally proved read-like and successfully invoked on a live, non-stale runtime owner?
Canonical result
Section titled “Canonical result”methods scanned 18,625zero-arg non-void 4,411capability records 3,021blocked 16,874 dangerous 69 unsupported 15,535IL-safe 1,751IL-rejected 1,270owner resolved 1,505no runtime owner 246invoked 1,505VERIFIED 1,462OBSERVED-THREW 43public VERIFIED 1,116reflection VERIFIED 346unsafe invoked 0stale invoked 0stability 1429 same / 33 changed / 0 failedengine exceptions 0Why the initial green run was rejected
Section titled “Why the initial green run was rejected”The first 0.3 run was structurally green but the first available Actor sample often came from ActorManager._dead_objects[]. That produced false OBSERVED-THREW/false-useful results for methods requiring a live actor. A second bug treated any method name containing attack as dangerous before considering that a method could still be a read-only query.
fix1 added sample-quality ranking and stale prevention; the canonical Actor sample moved to City._professions_dict{value}[] with quality 110. It also made read intent precede the broad dangerous-name heuristic.
Evidence semantics
Section titled “Evidence semantics”VERIFIED means successful call on the exact non-stale runtime sample after structural read-only proof. It does not mean a semantic guarantee for all possible world states.
Representative verified records now documented in detail include Actor, City, Kingdom, Building, WorldTile, MapBox and SaveManager.
Machine data: /worldbox-modding-docs/data/wbml/0300-capability-verification-atlas.json
Evidence: /worldbox-modding-docs/evidence/wbml-0300-result.txt
Detailed capability atlas · Actor reference · Next: WBML 0.4