# Gooo graph input assessment integration

Workbench PR #23 merged at 2026-10-08T07:02:27Z into main
`bc3f50f3f6e269888d1a047fb984feaec4327f82`. The tested branch head was
`b222d2a0c479daf727b4ad4b8ae1d87d3bdba3df`; its tree matches merged main.

Both PR run `37740588549` and push run `37740583747` succeeded at that head.
Immutable PR artifact `11532874066` (`gooo-ecosystem-observations`, 4,583,415 bytes)
matched GitHub's archive SHA256
`83ac6b71c6e6cfaadee270228cc4607b8b1278e9428a5953f7d162947227933c`.

The selected Linux CI records are preserved here. The input digest, explicit
feature/SDK versions, eight complete-array group hashes/memberships, counts,
model-observed false, zero calls, Gooo policy/source/compiler identity and replay
flags match the local source-bound observation. Runtime records show one supplied
observation passed, zero model calls and successful projection/runtime replay.
The assessment returns `input-consistent` / `evaluate-chooser`.

Fresh PR state was OPEN, non-draft, CLEAN and MERGEABLE; both check results were
successful and main was still `d7a628fabec94022e80fd6fcdf9b3580a3388b98` before
the normal expected-head squash merge. `merge.json` records the outcome.

No v3 weights were trained or measured. One filename-classifier family has eight
candidate arrangements; this report does not establish general prediction
quality or absence of collisions on other programs. Source/build identity and
original local observations remain in the workbench publication directory.