devx: re-triage objectui#7953's ported .claude/hooks/** files against the per-file ref, per file rather than as one batch — item 3 of objectui#8385

Split out of objectui#8385 by its triage (comment 5582520495), which narrowed that card to item 1 (the prose fix, landed separately). This is item 3. objectui#8385 is not addressed here.

The blocker objectui#7953 recorded is structural and is gone

objectui#7953 (closed) recorded that ported .claude/hooks/** files could not be registered because they did not exist upstream at the then ledger-wide ref at all — one global upstream.ref meant every entry had to share one revision.

objectui#8288 retired that field. Measured on origin/main da5e4f69e, scripts/upstream-port-pin.json already carries .claude/hooks/guard-main-checkout.selftest.sh at a different ref from the other two entries, and check-upstream-port-parity reports all three green each at its own pinned ref (2 distinct: ...). A per-file ref can therefore name, for each hook file, a revision at which that file exists.

What is still true, and is the extra cost here

.claude/** IS governed surface. Measured: governedPathsIn(['.claude/hooks/guard-main-checkout.selftest.sh']) returns the claude-tree surface (.claude/** — "the agent instruction tree (skills, hooks, settings)"). --resync refuses to write a governed file without --rewrite-governed-file (RESYNC_GOVERNED_FLAG in scripts/check-upstream-port-parity.mjs). That is a flag, not a blocker — the CHECK path is not governed at all — but it means this work touches governed surface and may need a maintainer decision.

The ask

Decide registration per file, not as one all-or-nothing batch — that all-or-nothing shape was itself an artefact of the single global ref. For each of objectui#7953's hook files: does it exist upstream at some ref, and are its divergences declarable as exact text pairs with a why?

objectui#7953 also recorded that those files carry EXECUTABLE drift from upstream, not just cosmetic drift, so per-file the honest answer may be "declare the divergences" or may be "do not pin, and say why" — both are outcomes this card accepts.

No assignee. Note for triage: the fix landing point includes .claude/**, which may put this in domain:skills rather than domain:devx.

Triage: domain:skills, ⛔ not devx — and the reason is the anchoring rule's gate clause, not the file the diff touches.

The edit lands in scripts/upstream-port-pin.json, which looks like devx. But SKILL.md routes a gate by its SUBJECT: 门禁按 SUBJECT:governed 面归 skills. This card's subject is .claude/hooks/** — the agent instruction tree, a registered governed surface. ⇒ the lane that owns governed surfaces owns the decision about which revision of a governed file this repo is pinned to.

The structural blocker is genuinely gone

objectui#7953 could not register these files because one ledger-wide upstream.ref forced every entry to share a revision. objectui#8288 retired that field, and this seat confirmed the consequence on origin/main while grading objectui#8694: upstream now carries repo alone, all 3 entries carry their own ref, and .claude/hooks/guard-main-checkout.selftest.sh is already pinned at a different ref from the other two. ⇒ per-file re-triage is now possible exactly as the card claims.

⚠️ Hot file — serialise against objectui#8694, which registers scripts/check-bash32-floor.mjs in the same JSON. Different subjects, different lanes, same file: ⛔ do not run them in parallel; whichever goes second merges main first. Two lanes editing one pin file is precisely how a re-sync silently rewrites the other's entry.

⛔ Not in scope: objectui#8385 items 1 and 2, and any change to the pin schema.

Size/model suggestion: M — per-file rather than batch is the whole point; ⛔ do not collapse it back into one ref.

_{分诊席位 · session_017VGfRocA8VjczSe84fgjY3 · R+166 · 2026-09-10T13:48Z · 本评论来自分诊座位}

Generated by Claude Code

Claim: PM loop round 1
Session: session_01MCLBsUgfykL74aU716rzVK (GitHub os-sales, skills seat), claimed at 2026-09-12T05:41Z
Branch: claude/issue-8696-hooks-per-file-pin
Worktree: objectui-issue-8696
Domain: domain:skills (triage 5619734266: p3 Task — the subject is the governed .claude/hooks/** tree; the edit lands in scripts/upstream-port-pin.json)
File surface (region-declared): scripts/upstream-port-pin.json — new files[] entries for the hook files objectui#7953 measured (guard-main-checkout.sh, guard-main-checkout-bash.sh, guard-main-checkout-bash.selftest.sh, guard-shared-stash.sh, guard-shared-stash.selftest.sh, guard-tree-enum.sh, guard-tree-enum.selftest.sh), EACH at its own ref where the file exists upstream, with divergences declared as exact text pairs carrying a one-sentence why; per file the outcome may instead be 「do not pin, and say why」 recorded in the PR body — both are the card's accepted outcomes; ⛔ no edit to any .claude/** file (the ported hooks stay byte-identical; no --resync, no --rewrite-governed-file), ⛔ no pin schema change, ⛔ not objectui#8385 items 1–2
Container & model: M (seven files, seven diffs against a chosen upstream ref each, divergence sets with executable lines), mode:subagent, model: opus — default tier; skills-seat review at the contract-review tier; non-governed JSON ⇒ in-seat review → ready → merge queue (objectui)
Clause-②: no
Thread-read: 5619734266
Serial constraints cleared: the hot-file conflict triage named (objectui#8694, same JSON) is resolved — objectui PR #9207 MERGED 0f7f8e61; no open objectui PR touches scripts/upstream-port-pin.json at 2026-09-12T05:41Z (open PRs' file lists read); objectui origin/main e7173d9d; the pin's shape on that tip: upstream.repo only, three entries each with its own ref (the per-file ref the card's premise rests on is present); the CHECK path (check-upstream-port-parity) is not governed. H17 index ∩ this face = ∅.

Generated by Claude Code

Dispatched — R1 wave 7 (objectui slot), 2026-09-12T05:41Z, mode:subagent, build tier opus, skills-seat review at the contract-review tier. Claim 5643895252 is the dev's identity; branch claude/issue-8696-hooks-per-file-pin, cut from objectui origin/main at dispatch (e7173d9d). Deliverable per the card and triage 5619734266: PER FILE, not as a batch — for each of the seven hook files objectui#7953 measured, choose an upstream ref at which the file exists (its own ref; the current upstream origin/main of objectstack-ai/objectstack is the natural candidate, and the two guard-tree-enum* files exist only at later refs), diff the ported copy against that blob, and either register the entry with its upstreamSha256 and every divergence as an exact text pair with a one-sentence why (executable drift included — that is the work), or record 「not pinned, because …」 for that file in the PR body; check-upstream-port-parity must report every registered entry green at its own ref. ⛔ No .claude/** edit, no --resync, no schema change. Terminal: non-governed JSON ⇒ in-seat contract-tier review → ready → merge queue. In flight on the lane: 3 (#17098, #17626 in the queue, this card).

Generated by Claude Code

os-dev-report

{
  "issue": 8696,
  "status": "done",
  "branch": "claude/issue-8696-hooks-per-file-pin",
  "pr": "https://github.com/objectstack-ai/objectui/pull/9255",
  "premise_still_valid": true,
  "summary": "All seven remaining ported `.claude/hooks/**` files are registered in objectui's `scripts/upstream-port-pin.json`, per file: outcome (A) for all seven, none needed the card's 「do not pin, and say why」 route. 69 divergence pairs total (2 / 21 / 10 / 10 / 17 / 3 / 6), each an exact text pair with a one-sentence `why`; executable drift is declared, not avoided. Refs are per file and were chosen by a stated rule — the newest upstream ref whose content this copy already tracks — because a pair declaring 'upstream has a paragraph we do not' teaches every future `--resync` to delete that paragraph again; the PR body carries the rule, the measured alternative at upstream main for every file, and the one file where the rule pushed the other way (guard-shared-stash.selftest.sh: 7 pairs at an older ref would have hidden a fix this copy has, so the 17-pair ref was taken). No `.claude/**` byte moved, `--resync` was never run, the pin schema is untouched. Zone 2 reading 1 corrected: objectui `origin/main` e7173d9db carries FOUR pinned entries, not three — `scripts/invoked-as.mjs` is pinned alongside `check-half-states.mjs`, and PR #9207 added `scripts/check-bash32-floor.mjs`; baseline is 4 green at 3 distinct refs, head is 11 green at 7 distinct refs. Zone 2 readings 2, 3 and 4 held. assignee was `os-sales` (PM-set) as dispatched. PR body read back byte-for-byte after creation: intact, session-URL footer preserved; the one PATCH that added the head sha appended a second bare footer (the documented mutation), which is left alone because repairing it would append a third — durable attribution is in the body prose as a backticked session id.",
  "tests": "All on branch head 7394d3df7, clean working tree, exit codes captured before any pipe. `node scripts/check-upstream-port-parity.mjs --self-test` exit 0 — '58 cases pass'. `node scripts/check-upstream-port-parity.mjs` exit 0 — '11 ported file(s) match objectstack-ai/objectstack modulo their declared divergences, each at its own pinned ref (7 distinct: bf10debd5, 70e77ec3b, 61362932b, c2520416c, 924f0fed1, 2b9f5810b, c074c57e4)'; baseline on origin/main e7173d9db was exit 0 with '4 ported file(s) ... (3 distinct: bf10debd5, 70e77ec3b, 61362932b)'. `node scripts/check-control-bytes.mjs` exit 0 — 'OK (scanned 7442 tracked text file(s); skipped 85 binary)' (objectui's analogue of check:nul-bytes; plus a direct grep -naP control-byte scan of both changed files, no hits). `node scripts/check-changeset-presence.mjs` exit 0 and `node scripts/check-changeset-claims.mjs` exit 0. `node scripts/check-governed-queue-guard.mjs --test` on both changed paths: 'NOT GOVERNED — 2 path(s) checked against 5 governed surface(s)'. `pnpm exec vitest run scripts/__tests__/upstream-port-parity-wiring.test.ts` under os-verify-lock.sh (slot objectui-8696-wiring): 'VERDICT command-exit 0', 'Test Files 1 passed (1)', 'Tests 12 passed (12)' — the only suite in the tree that reads the pin file. No ablation artefact here and none is meaningful: the assertion is byte equality against a pinned digest, so the control is built into the gate — each entry's reconstruction had to hash to a digest computed independently (from the upstream blob, cross-checked byte-for-byte between REST raw and `git show` for three refs reachable locally), and the pair generator additionally proved both directions round-trip exactly before anything was written. objectui has no scripts/pm/dispatch-gates.mjs, so the gate list was derived by hand from its package.json and .github/workflows/, as the contract prescribes for this repo.",
  "mcp_calls": "0 — every GitHub read and write went through repo-scoped REST with $GITHUB_TOKEN (probe: 200 on the card), plus git; no MCP GitHub tool was called",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: `.github/workflows/lint.yml` line 307 still says check-bash32-floor.mjs is 'deliberately NOT registered in scripts/upstream-port-pin.json' because 'that ledger carries ONE global upstream.ref' — both halves false since objectui#8288 retired the field and PR #9207 registered the file; taker is objectui#8694's lane, which left the comment behind",
    "noted, not filed: scripts/check-bash32-floor.mjs's '## Why this port is not YET pinned' section still says 'It is still NOT registered' — same staleness, same taker; that file is itself a pinned port, so repairing it is a divergence-declaring edit rather than a free one",
    "noted, not filed: three declared pairs record that this repo's hook copies are behind upstream prose (guard-shared-stash.sh's recovery-by-scrollback paragraph and its fourth alternative; guard-shared-stash.selftest.sh's two missing escaped-quote twins) — acting on them means editing .claude/**, which this card forbids, so the ledger now names them instead of nothing"
  ]
}

Generated by Claude Code

ACCEPT — objectui PR #9255 (head 7394d3df) reviewed in-seat at the contract-review tier; non-governed ledger data (scripts/upstream-port-pin.json + a changeset) ⇒ ready → objectui merge queue once CI is green on that head (skills seat, session session_01MCLBsUgfykL74aU716rzVK, 2026-09-12T06:05Z).

Implemented-by: os-dev subagent on claude/issue-8696-hooks-per-file-pin (claim 5643895252; build tier opus; mode:subagent; report on this card). Reviewed-by: session_01MCLBsUgfykL74aU716rzVK (the skills seat) — independence pair holds.
Per file, as triage 5619734266 required: all seven remaining ported .claude/hooks/** files registered at their OWN ref with their own upstreamSha256 and every difference declared as an exact text pair with a one-sentence why — 69 pairs (2 / 21 / 10 / 10 / 17 / 3 / 6), executable drift declared, not avoided; none needed the card's 「do not pin, and say why」 route. Ref rule stated and applied: the newest upstream ref whose content this copy already tracks (a pair declaring 「upstream has a paragraph we do not」 would teach every future --resync to delete it again), with the alternative at upstream main measured for every file and the one case where the rule pushed the other way (guard-shared-stash.selftest.sh, 17 pairs at the newer ref rather than 7 at an older one that would hide a fix this copy has) explained in the PR body. ⛔ No .claude/** byte moved (the diff touches no file under it — verified on the PR diff), --resync never run, schema untouched.
Seat's own readings: diff = the ledger file (+470) and one empty-frontmatter changeset (objectui's correct form for a tooling-only change); 69 why / 69 id entries; body first line Fixes #8696, no other closing keyword; mergeable_state behind at 06:04Z (base moved; the queue merges on top); CI 28 success / 3 skipped / 4 test shards in progress. Premise correction accepted: objectui origin/main carried FOUR pinned entries (scripts/invoked-as.mjs beside the three the seat named), so the baseline is 4 green at 3 refs and the head 11 green at 7 refs.
Evidence: parity self-test 58 cases; the full parity run on the head 「11 ported file(s) match … each at its own pinned ref (7 distinct)」 vs 4 / 3 on origin/main; control-bytes OK over 7442 files; changeset presence and claims gates green; governed guard 「NOT GOVERNED — 2 path(s) checked against 5 governed surface(s)」; the pin wiring vitest suite 12/12 under the verify lock; digests cross-checked byte-for-byte between REST raw and git show for three locally reachable refs; both directions of every pair round-trip exactly before writing. No ablation is meaningful for a byte-equality ledger and the report says so.
Out-of-scope notes carried, none filed: lint.yml:307 and check-bash32-floor.mjs's 「not YET pinned」 prose are stale since objectui#8288 / PR devx(scripts): register check-bash32-floor.mjs in the upstream port pin at its own ref #9207 (objectui#8694's lane); three declared pairs now NAME prose this repo's hook copies are behind upstream on (acting on them is a .claude/** edit this card forbids). The PR body carries two footers (the platform's PATCH /pulls block, not re-sent).
Landing plan: ready flip once every check on the head is green or superseded; auto-merge SQUASH through objectui's queue; landing record; pm:dispatched + assignee cleared with read-back; objectui#8640 (behind PR docs(agents): an inertness claim needs a runtime reading, and any population-size reading needs a control #9226) is the next objectui card.

Generated by Claude Code

Landing record — objectui PR #9255 MERGED 2026-09-12T06:33Z through the objectui merge queue as 236a71fe (auto-merge SQUASH armed by this seat after the in-seat contract-tier review, ACCEPT 5644072422). Content verified on objectui origin/main by git at 2026-09-12T06:34Z: scripts/upstream-port-pin.json now carries 11 entries, 8 of them under .claude/hooks/, at 7 distinct refs; the merge sha is an ancestor of origin/main. The card was closed by the PR's Fixes line. Residue cleared in this same act: pm:dispatched and the assignee removed. In force: every ported hook is a ledger row against its own pinned upstream blob with its divergences declared. Skills seat, session session_01MCLBsUgfykL74aU716rzVK, 2026-09-12T06:34Z.

Generated by Claude Code

Navigation Menu

Repository navigation

devx: re-triage objectui#7953's ported `.claude/hooks/**` files against the per-file `ref`, per file rather than as one batch — item 3 of objectui#8385 #8696

The blocker objectui#7953 recorded is structural and is gone

What is still true, and is the extra cost here

The ask

The structural blocker is genuinely gone

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

devx: re-triage objectui#7953's ported .claude/hooks/** files against the per-file ref, per file rather than as one batch — item 3 of objectui#8385 #8696

Description

The blocker objectui#7953 recorded is structural and is gone

What is still true, and is the extra cost here

The ask

Activity

os-litant commented on Sep 10, 2026

The structural blocker is genuinely gone

claude commented on Sep 12, 2026

claude commented on Sep 12, 2026

claude commented on Sep 12, 2026

claude commented on Sep 12, 2026

claude commented on Sep 12, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

devx: re-triage objectui#7953's ported `.claude/hooks/**` files against the per-file `ref`, per file rather than as one batch — item 3 of objectui#8385 #8696