Runtime capability atlases

WBML 0.2.0 — structural atlas

Canonical probe: 0.2.0-fix3, run 0fa6e20674fc.

Metric	Canonical value
Types	2,905
Members	38,553
Type ID collisions	0
Runtime concrete types	697
Runtime specialized types	29
Safe runtime reads	8,488
Runtime read failures	0
Relations	19,150
Unresolved relation endpoints	0
Research queue	15,079
Stability	4,988 same / 12 changed / 0 failed

0.2 is a read-only census and relationship map. Discovered methods/properties/constructors were not automatically invoked. The queue is research prioritization, not a safety verdict.

WBML 0.3.0 — zero-argument capability verification

Canonical probe: 0.3.0-fix1, run 82c23a7c8916.

Metric	Canonical value
Methods scanned	18,625
Zero-arg non-void	4,411
Capability records	3,021
IL-safe	1,751
Owner resolved	1,505
Invoked	1,505
VERIFIED	1,462
OBSERVED-THREW	43
Public Verified	1,116
Reflection Verified	346
Unsafe invoked	0
Stale invoked	0
Stability	1,429 same / 33 changed / 0 failed

The initial 0.3 run was structurally green but sampled dead/stale Actor objects and treated read-like names containing attack as dangerous. fix1 corrected owner-sample quality and moved read intent before the broad name heuristic.

Reading one record

Example canonical Actor record:

signature: getAge()->System.Int32
visibility: public
owner_source: City._professions_dict{value}[]
invocation_status: VERIFIED
result_kind: scalar
result_type: System.Int32
snapshot: 76
stability: same

76 is the value on that sample, not a property invariant.

Machine data: /worldbox-modding-docs/data/wbml/0200-automatic-api-atlas.json and /worldbox-modding-docs/data/wbml/0300-capability-verification-atlas.json.

Quick Docs · Actor details · WBML 0.2 research · WBML 0.3 research