VTOL wind exposure can now be selected at the observed front, cruise, back or landing phase, alongside the existing startup default. Native Buddy commands, the standalone CLI and the mission workspace use the same phase choices and report actual phase entry and wind onset. The fixed north-facing crosswind cases remain bounded limits during front/cruise/back; landing-only exposure completes their declared routes. Explicit wind-facing crosswind plans retain their qualified outcomes.
The model uses continuous projected-incidence v3 longitudinal forces and VTOL thrust/heading separation. The shared lift cascade explicitly preserves original Quad-X quaternion arithmetic so existing multirotor-derived incident recordings still replay byte-for-byte. Actual archived-source custody and managed-dataset migration tests reproduced the earlier regression, then passed after that compatibility fix. Strict source replay, integrity, admission and deadline checks remain unchanged.
The homepage and docs introduction now reflect the owned runtime and implemented camera adapter, expose experimental status, scope home simulator limits, prioritize the neutral-only runtime bench and improve keyboard/docs navigation.
Validation: lint, app/runtime/STX types, native docs and standalone CLI builds pass. The compatibility checks pass 61 tests across six actual files; integrated phase/CLI/workspace checks pass 66 tests across five files. The site checks pass 53 existing follow/camera tests; its 91 generated pages contain 6,748 checked internal links without broken destinations or anchors. Actual compiled browser QA passes desktop and 320 px layouts, skip-link focus, simulator target loss and docs navigation. Independent Spec and Standards reviews cover the individual slices and final integration. Representative Buddy/copied-CLI reports are compared byte-for-byte against the prior reviewed VTOL reports.
Full exact-head CI remains required before main publication. Earlier failed workflow archives remain retained; the separate Mac HTTP child-exit investigation remains open in #93. This remains fictional simulation and bench development, with physical commands disabled and board outputs inhibited.
Refs #87. Includes the pending runtime work from #91 and the site corrections from #94.