finding(devx): the other 7 ported `.claude/hooks/**` files all carry EXECUTABLE drift from the pinned upstream ref, and 2 of them do not exist at it — the ledger cannot reach them under one global `upstream.ref`

[baozhoutao]

Measured while landing the first .claude/hooks/** entry in scripts/upstream-port-pin.json. Part of #7263, which registered exactly one file and ruled the other hooks out of its own scope. Filed unassigned.

What was measured

Every file in .claude/hooks/ diffed against the upstream blob at the ref the pin already names, objectstack-ai/objectstack@bf10debd587f6ba891be9eadc2b76c91e15bd82b, and against upstream origin/main at a4816a79d. Differing lines are counted from a zero-context unified diff and split into executable and comment/blank by first non-space character. (The one row corrected after filing: this file vs upstream main, first written from a context-merged count.)

hook	vs the pin's ref bf10debd5	vs upstream main a4816a79d
guard-main-checkout.selftest.sh	REGISTERED by #7263 — 14 declared divergences	102 lines (40 exec / 62 comment)
guard-main-checkout.sh	67 lines (38 exec / 29 comment)	39 lines (19 exec / 20 comment)
guard-main-checkout-bash.sh	78 lines (8 exec / 70 comment)	111 lines (22 exec / 89 comment)
guard-main-checkout-bash.selftest.sh	17 lines (14 exec / 3 comment)	93 lines (62 exec / 31 comment)
guard-shared-stash.sh	101 lines (23 exec / 78 comment)	109 lines (22 exec / 87 comment)
guard-shared-stash.selftest.sh	61 lines (31 exec / 30 comment)	108 lines (50 exec / 58 comment)
guard-tree-enum.sh	absent upstream at that ref	17 lines (7 exec / 10 comment)
guard-tree-enum.selftest.sh	absent upstream at that ref	55 lines (37 exec / 18 comment)

Why this is the real work, and why it is not a bounded edit

1. Not one of them is a comment-only port. finding(devx): upstream-port-pin.json covers only scripts/, so the verbatim cross-repo ports under .claude/hooks/** have no drift gate — the same shape #6642 measured at ~212 KB #7263's body measured its one candidate at "exactly 4 declared divergences, all comment / echo-label text, no executable line differs". That was true of that file on the day it was measured and is true of no other hook here. Each remaining entry needs a divergence set whose text pairs cover executable code, with a one-sentence why per pair — which is the work finding(devx): upstream-port-pin.json covers only scripts/, so the verbatim cross-repo ports under .claude/hooks/** have no drift gate — the same shape #6642 measured at ~212 KB #7263 called "real work and the actual value", because writing it down is what turns "this line differs because someone decided so" from memory into ledger.

2. Two of them cannot be registered at all right now. guard-tree-enum.sh and its self-test do not exist in upstream's tree at bf10debd5; they arrived upstream later. The pin carries ONE upstream.ref for every entry (validatePin requires it to be a single 40-character commit sha, and --resync rewrites that one field), so registering them means either moving the pin's ref for all entries — which re-derives the three existing entries — or giving an entry its own ref. Both are schema-or-scope decisions, not implementation details, which is why this is filed rather than done.

3. A green ledger row is a statement about the PINNED blob, not about upstream today. The row finding(devx): upstream-port-pin.json covers only scripts/, so the verbatim cross-repo ports under .claude/hooks/** have no drift gate — the same shape #6642 measured at ~212 KB #7263 landed reads byte-identical to objectstack-ai/objectstack@bf10debd5 … modulo 14 declared divergence(s) and is exactly true. Against upstream main the same file differs by 98 further lines, two of which are live and already tracked: the slash-worktrees-slash substring hole (finding(hooks): guard-main-checkout.sh reads any git-dir path containing /worktrees/ as a linked worktree — a PRIMARY checkout under a dir named worktrees is unguarded, and by DEPTH #7259, PR fix(hooks): a linked worktree is one whose git-dir differs from its git-common-dir #7749 in flight) that upstream has closed, and the escape-hatch remedy sentence (hooks: the four guards' refusal messages advertise re-run with OS_ALLOW_*=1, a command prefix that cannot reach the hook's environment — twin of objectstack#15971 #7775, twin of objectstack#15971). Whoever takes this card should decide whether bumping the pin's ref toward current upstream comes first — that decides how much of the divergence work above survives.

Not asserted here

Whether the answer is per-entry refs, a ref bump, or registering only the hooks that are stable. #7263 ruled the scope question (the ledger MAY register governed files; --resync refuses to write them without a named flag) and deliberately left the enumeration open.

Filed by an automated developer session (Claude Code, PM round R44) while working #7263; no assignee, no priority label set.

---

Generated by Claude Code