[amaybaum]

Evidence landing under PR-AUDIT-1 (#1088), repair 5 of the removal-candidate audit (#1088 comment 6638207679).

Owner authorization (amaybaum, Claude Code session chat, 2026-11-04):

• "Composition and access: finite-depth composition contract, logical acceptance circuit and complete-record laws (research record) #912: Prepare a separate evidence-landing PR containing the four byte-identical review files. This retains the historical review record; it does not establish independent review of the theorem. No new scientific computation is authorized."
• After the governance: §A.46 prevent reuse of unverified premises and audit unnecessary evidence #1103 merge: "Composition and access: finite-depth composition contract, logical acceptance circuit and complete-record laws (research record) #912 evidence landing: Prepare a separate PR containing the four byte-identical review files, explicitly labelled as author-posted rather than independently authenticated." "Both PRs require independent review, CI, and separate merge approval." "No deletions, new scientific computations, or changes to frozen records are authorized."

Head

2bd2dee5a60488a8e96c7d7660118def98e1e741

• afa340c3 (parent fb117893, current main): the landing.
• 2bd2dee5: the exact-head review of afa340c3 (ACCEPT WITH CHANGES), applied to SCOPE.md only: R1 (name the supporting files; "the table above" had become ambiguous), R2 (record that two of REVIEW.md's statements rest on commits off main), and the optional note on the first table.

What lands

In verification/programmes/oi-qm/reconstruction/composition-access-20261024/review/, byte-identical to #912's branch at 2af1d5d5 (copied with git checkout 2af1d5d5 -- <path>; sha256sum -c SHA256SUMS passes in review/):

File	Blob
REVIEW.md	eceb7060ad337baa2e2161f6e9e0ba58d0c2ac2e
SHA256SUMS	c9a27deaba09519aa5a4e50dac3a79cf9574d7ae
independent-result.json	01cd98056d3d5b0f60fa6d16595d3ee01a3ef1a0
independent_checks.py	fb0aac3f4f2a1ea83bdd18d02abe9e84e9729c09

independent_checks.py uses the Python standard library only and reads no file. It was not run for this landing.

The one edited file: SCOPE.md

• review/ is removed from the list of Composition and access: finite-depth composition contract, logical acceptance circuit and complete-record laws (research record) #912 content not on main; of README.md's replay commands, the first two now read only files on main.
• A blob table for review/ is added.
• The review record is described as author-posted: committed from the author's account (commit 5b57321a), naming no reviewer, and not an independently authenticated review. It does not make REPORT.md's theorem independently reviewed; REPORT.md's own status ("Separate-author review remains pending") is quoted unchanged.
• It records that the reviewed author commit a7f22c38's REPORT.md, checks.py and RESULT.json have the same blobs as the copies on main, and that REVIEW.md's statements about 34b631c4 and 2d555f17 rest on commits not on main.

No file is deleted, no frozen record is edited, and no status of the theorem, of MR-1's Proposition B2 or of any manuscript claim changes.

Commit references in review/REVIEW.md (§A.42 item 8)

Reference	On main?	Role
a7f22c38	No	Support on main by content: the verdict is about this commit's REPORT.md, checks.py and RESULT.json, whose blobs equal main's (2f77fe2f, 8203bdcb, c4fc742e). Its README.md (e30a039c) and SHA256SUMS (f7053197) differ from main's (0c0d8b25, 2e85c15a).
34b631c4	No	Support off main (a gap): REVIEW.md:12–14 says the reviewed child changes only REPORT.md and its manifest. True (a7f22c38's parent is 34b631c4; their diff is exactly REPORT.md and SHA256SUMS), but it rests on off-main commits; SCOPE.md states this.
2d555f17	No	Support off main (a gap): REVIEW.md:84–88's "preserves their requirements" uses #885's acceptance rule at this commit, absent from main; SCOPE.md states this.
5f6786d6	Yes	On main.

PRs #912, #885 and #910 named in REVIEW.md are provenance. No reference is offered as support for a claim on main beyond the existence and content of this author-posted record.

Local checks at the head

legacy_records_check.py: 303 records intact. citation_roles_check.py, duplicate_check, staleness_check: pass.

Still needed before merge

Delta review of 2bd2dee5, CI at the exact head, and separate owner merge approval of the reviewed SHA.

🤖 Generated with Claude Code

https://claude.ai/code/session_013T6pfKuv5mi3MXtKjH2tZb

[amaybaum]

Merged. Merge commit e0f522322bee6ddc6f9d57e8f91d4f1aded64ffd, pinned to head 2bd2dee5a60488a8e96c7d7660118def98e1e741.

Owner approval (amaybaum, Claude Code session chat, 2026-11-04): "I approve the normal merge of PR #1105 at exact SHA: 2bd2dee5a60488a8e96c7d7660118def98e1e741". The owner's stated limitation: "This merge preserves the historical review evidence. It does not establish that #912's theorem was independently reviewed or verified. The unresolved off-main dependencies remain explicitly documented. No scientific status is promoted."

Immediately before the merge:

• The remote head was 2bd2dee5 and the base was main fb117893; GitHub reported the PR mergeable and clean.
• Exact-head review: ACCEPT WITH CHANGES at afa340c3, applied in 2bd2dee5; delta review ACCEPT at 2bd2dee5.
• CI run 41891430875 at the head concluded success: 25 checks succeeded and 1 was skipped by design; in Mathlib bridge, Build and Release gate succeeded.

After the merge:

• The parents are fb117893 and 2bd2dee5, and the tree (83054e00) equals the head's tree.
• git diff fb117893 e0f52232 changes 5 files in composition-access-20261024/: SCOPE.md and the four added review/ files.
• The four review/ files have the same blobs as at 2af1d5d5, and sha256sum -c SHA256SUMS passes for them.
• main CI on e0f52232: push run 41893327049 concluded success; 25 checks succeeded and 1 was skipped by design. In the Mathlib bridge job, Build and Release gate both succeeded.

---

Generated by Claude Code