#300 ADAPTER ASSESSMENT ROSTER — STATUS, as of 2026-09-12 (todlando).

THIS DIRECTORY IS NOT A COMPLETE ROSTER. Written because I read these same files and reported
"both adapter answers are in" to doyle, which was FALSE. doyle corrected it the same hour.

  claude-spt  — ANSWERED by perri, the adapter OWNER.
                perri-claude-spt-2026-09-12.txt (233056d8) + addendum (57ca1644).
                perri's co-authorship objection remains OPEN.

  omp-spt     — OUTSTANDING. NO assessment exists in this directory or anywhere else.
                Re-scoped request sent to emphasys 2026-09-12T10:51Z (QUEUED, emphasys offline).

  emphasys-harness-level-2026-09-12.txt (df7cac0c) is NOT the omp-spt answer. Its author states
  so themselves, unprompted, in its first paragraph: they do not own claude-spt, they are a
  Claude Code session RUNNING on claude-spt, their answer is harness-level evidence feeding
  perri's, and perri outranks them on adapter behaviour. It is SUPPLEMENTARY CLAUDE CODE
  EVIDENCE. Counting it as a second adapter verdict is the error this file exists to prevent.

NOT SETTLED, per doyle's ruling — do not read any of these out of the files above as fact:
  - "always fires" (conflicts with perri's reported bypasses and observation limits)
  - enqueue-versus-dequeue ordering (capture+report in one hook does not establish it)
  - absence of later prompt rewriting (no rewrite field in one schema does not prove it)
  - the 27-vs-26 hook count (DOCUMENTATION UNCERTAINTY in both directions, not proof of a
    missing input hook, and not proof the unaccounted event is input-related)
  - no all-or-nothing refusal ruling has been made; Reavo's submitter decision pends via lia.

Contract these were read against: docs/INPUT-PROVENANCE-CONTRACT.md at
d142aa740cf0cca35a553e35d76fcd64f88b1dae0659f55764b9b497e655d0a7 (re-hashed 2026-09-12, unchanged).