What
PR #614's work is not on master. It was merged into fix/names-and-docs-that-lie — PR #601's branch — which had already been squash-merged. It needs to be recovered onto master.
Verified facts
Root cause is filed separately (merge polls do not show the base branch at tap time; the merge grammar forbids --delete-branch, so #601's branch survived and GitHub fired no retarget).
What the work is
#614 = "Give D1 the Scryfall cross-check it advertised" — owner ruling 8, 2026-07-29: implement the real Scryfall printing-count check (issue #600, option b), counted from the bulk file already parsed. Zero new HTTP requests.
This matters because of what #596 and #600 established: the D1 gate's "cross-verified against Scryfall" condition has never excluded anything — printings_count is a count over our own rows, so the tier's first condition went 137 → 137. The gate has been advertising a verification it does not perform.
Migration
0102 is allocated to this recovery (scryfall printing count). Master's leaf is 0099; 0100 → #604, 0101 → #615. PR #611's migration-leaf guard is now on master and will fail CI on a graph that forks as merged, so the dependency must be set correctly at the time of the rebase, not fixed afterwards.
Status — RE-DISPATCH NEEDED, and STATE UNVERIFIED
Recovery was dispatched. Its agent died to an API error mid-run — last observed output "Now let me run the test suite". At handoff the worktree .claude/worktrees/agent-614-recovery state was UNVERIFIED: the Bash tool's safety classifier went down before it could be checked, so it is not known whether a branch was pushed or a PR opened.
FIRST ACTIONS:
- Check
.claude/worktrees/agent-614-recovery — git -C <worktree> status and git -C <worktree> diff.
- Check whether a PR was already opened (
gh pr list), so the recovery is not done twice.
- Only then cherry-pick
c22fc90e (or git diff a7eb00ab c22fc90e) onto a fresh branch off master.
Standing constraints: own worktree per agent (OPS-CORR-0007); never two agents on one branch; merges are owner-only via Signal poll and the poll's scope text must state the base branch.
What would close it
A PR based on master, carrying #614's 10 files / +557/-117, migration numbered 0102 with the correct dependency, tests green on a non-stale run, merged onto master; and scryfall_default_cards_printings_count present on master. Then close #600.
Refs
What
PR #614's work is not on master. It was merged into
fix/names-and-docs-that-lie— PR #601's branch — which had already been squash-merged. It needs to be recovered onto master.Verified facts
scryfall_default_cards_printings_count.state=MERGED,base=fix/names-and-docs-that-lie,mergeCommit=c22fc90e.git diff a7eb00ab c22fc90e= 10 files, +557/-117 — exactly D1: implement the Scryfall printing-count cross-check it advertised (count the bulk file we already parse) #614's work, and cleanly cherry-pickable onto master.Root cause is filed separately (merge polls do not show the base branch at tap time; the merge grammar forbids
--delete-branch, so #601's branch survived and GitHub fired no retarget).What the work is
#614 = "Give D1 the Scryfall cross-check it advertised" — owner ruling 8, 2026-07-29: implement the real Scryfall printing-count check (issue #600, option b), counted from the bulk file already parsed. Zero new HTTP requests.
This matters because of what #596 and #600 established: the D1 gate's "cross-verified against Scryfall" condition has never excluded anything —
printings_countis a count over our own rows, so the tier's first condition went 137 → 137. The gate has been advertising a verification it does not perform.Migration
0102 is allocated to this recovery (scryfall printing count). Master's leaf is 0099; 0100 → #604, 0101 → #615. PR #611's migration-leaf guard is now on master and will fail CI on a graph that forks as merged, so the dependency must be set correctly at the time of the rebase, not fixed afterwards.
Status — RE-DISPATCH NEEDED, and STATE UNVERIFIED
Recovery was dispatched. Its agent died to an API error mid-run — last observed output "Now let me run the test suite". At handoff the worktree
.claude/worktrees/agent-614-recoverystate was UNVERIFIED: the Bash tool's safety classifier went down before it could be checked, so it is not known whether a branch was pushed or a PR opened.FIRST ACTIONS:
.claude/worktrees/agent-614-recovery—git -C <worktree> statusandgit -C <worktree> diff.gh pr list), so the recovery is not done twice.c22fc90e(orgit diff a7eb00ab c22fc90e) onto a fresh branch off master.Standing constraints: own worktree per agent (OPS-CORR-0007); never two agents on one branch; merges are owner-only via Signal poll and the poll's scope text must state the base branch.
What would close it
A PR based on master, carrying #614's 10 files / +557/-117, migration numbered 0102 with the correct dependency, tests green on a non-stale run, merged onto master; and
scryfall_default_cards_printings_countpresent on master. Then close #600.Refs
mergeCommit=c22fc90e), PR printings_count counts our own rows: rename it, and delete the "cross-verified against Scryfall" claim #601, PR Fail CI on a migration graph that forks WHEN MERGED, not just on the branch #611. Issues The D1 'cross-verified against Scryfall' gate has never excluded anything: printings_count is a count over our own rows #596, deductive backfill: the first tier's "cross-verified against Scryfall" condition excludes nothing (137 -> 137) #600. Owner ruling 8, 2026-07-29.