PR #653 follow-ups: the predicate prose moved outside test_doc_pointers.py, and the comment saying it was filed was not nf-5pp ← Beads

Twinned from https://github.com/cfsh/ninefold/issues/695 by tools/beads/import_github.py's reconcile pass.

Suggestions from ninefold-reviewer's review of #653.

#653 merged (7718728d). Both items below are on main now. Unclaimed and unlabelled, per CLAUDE.md. #655 was the earlier follow-up issue for this PR and is closed, so this is a new one rather than an update.

1. The Godot-build prose now lives outside CURRENT_TREE_DOCS, and the comment saying it was filed was not

tools/tests/test_doc_pointers.py — the block under CURRENT_TREE_DOCS names tools/nflib/godot_build.py as a deliberate absence and closes with *"Filed rather than worked around."*

Nothing was filed. Searched open and closed issues by keyword, and walked every issue opened after the rebase commit (521df6c8, 02:23:18Z): #682–#691 are all unrelated. The commit message is more careful (*"left as a comment naming it"*); the file is the copy that outlives the thread, and it is the one that overclaims. This issue is the filing — put its number in the comment.

The gap itself. #637 moved the predicate and its dense prose out of tools/pr/rebuild_f5.py into tools/nflib/godot_build.py. CURRENT_TREE_DOCS cannot follow it: nflib/** is a governed surface and test_doc_pointers.py is an open perimeter test, so #633's test_no_open_test_file_guards_a_governed_surface refuses the quoted literal. The guard #606 §8 built for exactly the CsprojPremiseTest -> BuildPremiseTest rename therefore no longer scans the file that prose lives in.

Harmless today and worth saying so: godot_build.py carries no path:LINE citation, and notes/f5.md still cites BuildPremiseTest, so the rename case is pinned by proxy. What is missing is any check on the prose that moved.

Routes, none of them free: - Make test_doc_pointers.py governed (deny it in tools/nflib/perimeter.py) and then add the path. Correct, and the honest reading of the contract's own remedy — MENTIONS is for a file that merely *names* a governed path, which this would not be. - Split the doc set, with a governed test owning the governed half. - Accept the gap, and say so in the comment without claiming a filing.

The MENTIONS route is the one to avoid: it would exempt a file that genuinely does guard the surface, which is #674's shape.

2. #653's body is stale in four places, and it is now a merged PR's durable record

Advisory; a five-minute edit if anyone is passing.

• "Merging needs @aedanpope's approving review and he is offline ~8h, so this stops at *reviewed and green*" — approved 01:08:32Z, merged 02:32:46Z.
• "Review round 1 (Tier-0) is in and addressed" — two rounds landed, plus a re-stamp comment, then the approval.
• The body never mentions the rebase onto 900bc4d, and tools/tests/test_doc_pointers.py (+21/-1) is in the file list with no Test Plan entry — including item 1 above, which is the most consequential thing in that commit.
• The churn cell "1 of the last 39 merges" still sits in a column whose other cells (43, 9) are touch-counts over 200 commits. Over the same window that row is ~36. ⚠ Correcting the earlier review's framing: this does *not* leak into the commit message — 7718728d is the three commit messages concatenated, not the PR body.