Seven 2026-08-24 merges lack a clean review verdict and were not covered by the #645 revert

[dem-extra1]

A sweep of every PR merged into main on 2026-08-24 (UTC), classifying each one's latest review verdict at merge time
(via check-pr-fully-clean.py's classify_verdict, then hand-confirming each flagged comment's last ### Verdict heading),
found that #645 reverted only the #622-#633 batch.
Seven other merges from the same day lacked a clean review verdict and remain on main:

Merged with no review verdict at all

Each of these edits workflow files, so the claude-code-action guard posted
"No review ran --- this PR edits .github/workflows/..." and no reviewer ever produced a verdict.
No self-review fallback (per self-review-fallback.md) was posted on any of them before merge.

PR	merged (UTC)	evidence
#593	07:39:51	guard warning 00:36; no other review item
#594	07:40:11	guard warning 00:40; no other review item
#595	07:40:25	guard warning 00:47; no other review item
#596	07:40:43	guard warning 00:51; no other review item
#597	07:41:11	guard warning x3; OpenCode review failed x3; only Bugbot "found no new issues" (06:27)
#603	07:51:07	guard warning 07:50; merged ~3 min after opening

Merged over a standing Needs-more-work verdict (borderline)

• fix(workflows): remove invalid timeout-minutes on caller jobs (#582) #583 (merged 00:28:35): the last Claude verdict before merge
  (comment 5382332775, 2026-08-22)
  was Needs more work --- the residual finding being seven other workflows carrying the same pre-existing bug, out of this PR's diff.
  A local Antigravity review approved on 08-23 and a local adversarial review filed the remaining findings as issues (a Defer),
  but no subsequent clean Claude verdict exists.

The revert PR itself

• revert: revert PRs #622 through #633 #645 (merged 23:18:54) also merged with no review (same workflow-edit guard, merged 5 min after opening).
  It is the remediation --- a pure revert restoring main to 173e473 --- so its risk profile differs,
  but it is noted here for completeness.

Suggested disposition

Either revert these as #645 did for #622-#633, or run a retroactive review round on each
(a workflow_dispatch review, or the self-review fallback that was owed at merge time)
and fix or file whatever it finds.
The workflow-edit guard means these PRs could never get an in-PR Claude review by construction,
which makes the self-review fallback the mandatory step that was skipped.

Sweep performed 2026-08-24 ~16:30 PT; sibling ai-config findings are tracked in Morrison-Lab/ai-config#2174 and a parallel ai-config issue.