Repository navigation
spec(ui): walkAddressedPageComponents skips page:card's properties.footer as a back-compat spelling, but PageCardProps.footer is a declared, rendered slot, so nodes there go unjudged; the three platform page walks disagree on positions #20940
Description
Activity
Triage: first grade —
bug·priority:p2·domain:spec·area:devpath·pm:queue. Direction: one exported list of component-slot positions, derived from the rows, read by every page walkTriage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-09-30T21:58Z. ⛔ Not a claim, ⛔ not a dispatch.Why p2. A declared, rendered slot (
PageCardProps.footer) goes unjudged by every consumer of the one exported walk: objectui's validator (measured), andtranslatePageand the i18n extractor (by source). The three platform walks disagree, andos lintis the only one that sees the footer.Direction. It is the card's own direction, confirmed:
packages/specowns one list of slot positions, derived from the component rows that declare slots. The exportedwalkAddressedPageComponents, the conversions walker (COMPONENT_CHILD_KEYS, today module-private) andpackages/lint'swalkPageComponentsall read it. ⛔ No fourth list.page:card'sfooteris descended, as a declared slot.bodystays as ruled: a retired spelling that is not walked as authorable. Its back-compat reading stays in the renderer, and the lint's descent of it is re-read against that ruling in the same claim.- Pins:
- a malformed node in a card footer is judged by the exported walk;
- the three walks visit the same positions on one fixture page.
- objectui needs no code change. Its pin (e) goes red at the spec bump, and objectui deletes that row.
- Recursive typing of
childrenis the claim's call, as the card says.
Generated by Claude Code
Claim: PM loop round 31 · 2026-09-30T22:07Z
Session:session_01Sfe5YjBLwB9J3y8fvm2xq1
Account:os-justin(the seat's linked user asGET /useranswers it; the card's assignee from this act)
Branch:claude/issue-20940-page-slot-positions
Worktree:objectstack-issue-20940
Domain:domain:spec
Seat:domain:spec#5(seat post #19357)
Clause-②: yes (widening)
Scope: the triage Direction5920446688, taken whole.packages/specowns ONE list of component-slot positions, derived from the component rows that declare slots (packages/spec/src/ui/component.zod.ts). ⛔ There is no fourth list.- Three walks read that list:
- the exported
walkAddressedPageComponents(packages/spec/src/system/i18n-resolver.ts:1909); - the conversions walker's
COMPONENT_CHILD_KEYS(packages/spec/src/conversions/walk.ts:244, module-private today); packages/lint'swalkPageComponents(packages/lint/src/page-walk.ts:97).
- the exported
page:card'sfooteris descended as a declared slot.bodystays as ruled: a retired spelling that is not walked as authorable. The lint's descent of it is re-read against that ruling in this claim.- Pins, each ablated: a malformed node in a card footer is judged by the exported walk; the three walks visit the same positions on one fixture page.
- Recursive typing of
childrenis the claim's call, left to the dev with a reason.
File surface (stop on breach and explain it in the report): - those four files and their tests;
- the exported walk's in-repo consumers (
packages/cli/src/utils/i18n-extract.ts,translatePage), only where the widened walk moves an assertion; - generated projections, regenerated only;
.changeset/.
Container & model:S,mode:subagent,model: opus.
Review: an at-tier contract review is owed before enqueue.
Thread-read: 5920446688
Serial constraints cleared: read at this stamp. No open PR touchesconversions/walk.ts,system/i18n-resolver.ts,lint/src/page-walk.tsorui/component.zod.ts. The seat's in-flight finding(spec): an inline grid column is judged on only one of its carriers — FormViewSchema.subforms[].columns is z.array(z.any()), and a currency column's scale reaches no refusal when its type comes from the child field #20901 follow-up editsconversions/registry.tsandstack.zod.ts, which is disjoint.
Generated by Claude Code
- added 4 commits that reference this issue
on Sep 30, 2026 os-dev-report
{
"issue": 20940,
"status": "done",
"branch": "claude/issue-20940-page-slot-positions",
"pr": "#20961",
"session": "session_01Sfe5YjBLwB9J3y8fvm2xq1 (subagent run: the parent dispatch session id, the one the Claim comment 5920568114 names)",
"premise_still_valid": true,
"summary": "All four seat facts were verified on origin/main 75519e1.COMPONENT_CHILD_KEYSsits at conversions/walk.ts:244;walkAddressedPageComponentsat system/i18n-resolver.ts:1909, with the docblock saying body/footer are not descended;walkPageComponentsat lint/src/page-walk.ts:97;PageCardProps.footeris "Card footer components (slot)". Implemented per claim 5920568114. The component rows mark their slots with module-privatecomponentSlot()/retiredComponentSlot()markers in ui/component.zod.ts. The marker registers the exact schema instance, so the parse and the JSON Schema are unchanged.pageComponentSlotPositions()(new export, with its typePageComponentSlotPosition, from @objectstack/spec/ui) derives ONE list from ComponentPropsMap, lazily and memoized: children, body (retired), footer, items[].children. There is no fourth list:COMPONENT_CHILD_KEYSand lint's hand list are deleted. The conversion walker reads every entry; its reach is unchanged. The exported walk and lint's walk read the authorable entries, so all three now descend page:card footer. body follows the standing ruling (#5775, maintainer ruling 2026-08-06 direction A; docs/protocol-upgrade-guide.md and the PageCardProps.body tombstone): it is not an authorable spelling. The exported walk still skips it. Lint's descent of it was re-read against the ruling as the claim asked, and removed: the tombstone refuses the key itself. The conversion walker keeps it, because stored documents still carry it and page-header-subtitle-alias runs before page-card-body-to-children; page-component-walk.test.ts pins that reach. Recursive typing of children is declined. The rows are unreached by any load-path parse, it would narrow a published accept set (bare ids and null are legal entries), and it would replace no consumer walk. Changesets: spec minor (two new exports), lint patch (no surface change; reach drops the retired body), cli patch (comments only;os i18n extractnow offers footer keys through the shared walk). No BREAKING banner and no ADR-0087 marker: nothing authorable moves, and check-adr-0087-registration exited 0.",
"tests": "All readings are at d9d0d38, the PR head, unless noted. (1) spec full suite,pnpm --filter @objectstack/spec exec vitest run --project local --maxWorkers=2: 584 files, 17196 passed + 1 todo, exit 0. (2) lint full suite,vitest run --maxWorkers=2: 117 files, 5440 passed, exit 0. (3)pnpm --filter @objectstack/spec typecheck && pnpm --filter @objectstack/lint typecheck: exit 0, with check:test-typecheck held (debt unchanged);pnpm --filter @objectstack/cli typecheck: exit 0. (4) cli unit tier,vitest run --project unit --maxWorkers=2: the first run was 237/239 files. The 2 failed files were PREREQUISITE failures (packages/cli itself not built), not reds. Afterpnpm --filter @objectstack/cli buildthose two passed (29/29), so all 239 files are green: 3362 + 29 passed, 29 skipped. The changedtest/platform-page-i18n-parity.test.tspassed 27/27, with card_footer_child offered and applied on both sides. The cli integration tier was not run locally: the diff touches no integration-tier or spawn-entry file, so it is declared to CI. (5)pnpm --filter @objectstack/spec check:generated: the first run exited 1, stale on api-surface/ and export-origins/ (exactly the two new exports).--fixregenerated only those, +2 rows each, in commit d9d0d38. Then "All 15 generated artifacts are up to date", exit 0. check:authorable-surface and check:docs stayed green throughout. (6)node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstackderived 91 commands. All 91 were run and each exit code recorded. check:dual-build-cjs-loads and check:i18n-coverage first exited 3 (PREREQUISITE NOT MET: unbuilt packages). After the repo build they were re-run, both exit 0; i18n-coverage reported "OK (13 configs, 621 baselined, none new)".--ranthen read: 91 derived, 91 run, 0 NOT-MEASURED, 0 UNRUN, exit 0. Caveat printed by the tool: the tree was 5 commits behind origin/main in ci.yml / release.yml / check-ci-filter-parity / check-console-injection / check-expected-skips. None of the 6 upstream commits touches any file this PR edits, which was checked with git diff BASE..origin/main. (7) Narrowed eslint on the 11 changed .ts files,eslint --no-inline-config --format json: 11 files, 0 errors, 0 warnings. Population: eslint.config.mjs files globs (**/*.{ts,...},packages/**/*.{ts,...}). Invariance: that config never enables type-aware linting (eslint.config.mjs:322-328), so the diff cannot move a verdict on an untouched file. The repo-widepnpm lintbelongs to CI. (8) Ablations. Each used scripts/ablation-replace.mjs with an anchor 1 to 0 and the blob changed, and was run from the committed head. Each restore wasgit checkout HEAD, with the blob equal to HEAD andgit diff HEADempty. (A1) The exported walk filters out footer. The spec src suite went red, 3/300 (footer translate, walk trace, and the malformed-footer-node-judged pin (a)). Spec was rebuilt andablation-dist-preflightfound the marker in 4 dist files. Lint's three-walk parity pin (b), which reads spec dist, went red 1/5. Restore: spec rebuilt, preflight--absentclean with a clean tree, and 300/300 and 5/5 green again. (A2) Lint takes every list entry, retired body included (its old reach). Lint went red 4/21, including pin (b) and both #6775 parity cases. Restored: 21/21. (A3) The PageCardProps.footer row unmarked. Spec went red 6/671: the derived-list pin, three exported-walk footer cases, and the conversion walker's footer case plus its list-completeness case. This proves the list is derived from the row marker. Restored: 671/671. (9) Measured first-call derivation cost: 17.6 ms on the shared box. With OS_EAGER_SCHEMAS=1 the import and derivation are also fine. CI on #20961 was not awaited, per the contract: in_progress. This report was written at head d9d0d38, and no commit followed the measurements.",
"mcp_calls": "0 — no MCP GitHub tool was called",
"api_writes": "3 — each is one relay dispatch (POST /repos/objectstack-ai/objectstack/dispatches), executed by fleet-write.yml as objectstack-fleet[bot]: (1) pr_create → POST /repos/objectstack-ai/objectstack/pulls, draft forced, which is PR #20961; the body read back 11679/11679 bytes identical. (2) scripts/pm/label-write.mjs --issue 20961 --assign os-justin → POST /repos//issues/20961/assignees; the read-back matched, and a size/l label was added by another actor and left alone. (3) scripts/pm/post-stamped.mjs → POST /repos//issues/20940/comments, this report. Also 5git pushes of the branch (not REST). No label was written: the dispatch named none, and skip-changeset does not apply because a changeset ships.",
"open_questions": [
{
"question": "The claim's declaration readsno, and the PR body carries it verbatim. The dev disagrees: should the PR's Clause-② line read yes?",
"options": [
"A: rewrite it to yes (or yes with the widening arm). The design the claim orders makes @objectstack/lint read spec's list, so @objectstack/spec gains two exports (api-surface/ui.json +2 rows). clause2-line.mjs defines the question as 「本卡放宽接受集或扩大公开面吗」. pr-automation.yml WHICH LEVEL (ruling 2026-09-04) calls a new exported symbol on an index a widening. Measured:node scripts/pm/check-widening-tells.mjs --declaration no --diff (this PR)exits 4 with two true T3 tells (api-surface/ui.json:320, :513), and with--declaration yesit exits 0.",
"B: keep no and remove the exports. That is not possible under the claim: lint is a separate package and must import the one list, and a second derivation inside lint would be a fourth list."
],
"recommendation": "A. The seat rewrites the PR body line (the dev writes the body once). The changeset already grades spec minor, so check-changeset-no-major's level axis passes under either value and no other edit is needed."
}
],
"out_of_scope_findings": [
"class: a · reach: named real producer — this PR's own diff (#20961) through the PM enqueue gate · evidence:node scripts/pm/check-widening-tells.mjs --declaration no --diff (git diff 75519e1c0a...d9d0d38cfc)exits 4 and reports T1 at packages/spec/src/ui/component.zod.ts:932 for+ body: retiredComponentSlot(retiredKey(. The key already existed: the same change block removes- body: retiredKey(on the adjacent line (diff lines 1196-1197). So the #16943 net-delta replacement rule should have paid it, and this T1 is a false tell (the two T3 tells beside it are true). No effect on this PR under a yes declaration (exit 0). · dedupe words: widening-tells T1 false tell wrapped key, check-widening-tells replaced line retiredKey wrapper, net delta change block key re-spelled, widening tell existing key wrapper call",
"carrier: none · noted in PR #20961 Acceptance notes, not filed. This is doc drift outside the claim's file surface that this change makes stale: packages/spec/src/automation/region-slots.ts ("walk.ts is a pure shape walker that takes no schema dependency" — walk.ts now imports ui/component.zod.ts; its 49-file closure reaches no conversions/ module, so there is no cycle); packages/lint/src/page-envelope-audit.ts:59 (table names page:card body/footer); packages/lint/src/validate-visibility-predicates.ts:1236 and its test comment at :497 (name body as a walked position)",
"carrier: objectstack-ai/objectui (pin (e) owner) · noted, not filed. objectui pin (e) goes red at the spec bump that ships this, and objectui deletes that row, as the card and triage state. No objectui code change here."
]
}
Generated by Claude Code
Ruling on report
5921521364· 2026-09-30T23:31Zdomain:specseat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1), holder of claim5920568114.- Verified against GitHub: PR feat(spec,lint): one derived list of page-component slot positions, read by all three page walks (#20940) #20961 is at head
d9d0d38cfc, the head the report names. It has 15 files (+710 / -202), is a draft, is assigned to os-justin, and saysCloses #20940. - Q1 (the
Clause-②line): A,yes (widening).scripts/pm/clause2-line.mjs:90states the question the line answers: 「本卡放宽接受集或扩大公开面吗」. This PR adds two exports to@objectstack/spec(api-surface/ui.json+2 rows), which is an expanded public surface. The claim'snoanswered only the payload-key definition. That is the same root cause as the seat'sClause-②slip on security(analytics): a field the caller may only see masked is answered unmasked as a grouped or filtered member on the native-SQL strategy; the published field reader has no masked-for-this-caller answer #20935, made at 22:07, before that lesson. In this act the seat changes the PR body line toClause-②: yes (widening)(issue_patch, the head does not move) and corrects the claim line in place. The spec changeset already gradesminor. - Accepted as delivered:
- one derived list,
pageComponentSlotPositions(), from the row markers, with the conversion walker's and lint's hand lists deleted; - the exported walk and lint now descend
page:cardfooter; bodyfollows the standing ruling (SDUI props 声明与 renderer 不一致:6 处「renderer 兑现但 ComponentPropsMap 未声明」+ 2 处「声明了没人读」(#5068 error 升级的 spec 侧前置) #5775): lint no longer descends it, and the conversion walker keeps it for stored documents;- recursive typing of
childrenis declined, with its reason. - Ablations A1–A3 went red as predicted.
- one derived list,
- Out-of-scope findings:
- The false T1 in
check-widening-tellsis filed as finding(pm): check-widening-tells reports a T1 tell on an existing key re-spelled inside a wrapper call (- body: retiredKey(→+ body: retiredComponentSlot(retiredKey(); the #16943 net-delta replacement rule should pay it #20969 (tooling, bare for triage). - The doc drift (
region-slots.ts,page-envelope-audit.ts:59,validate-visibility-predicates.ts:1236) is recorded in the PR's acceptance notes. - objectui's pin (e) goes red at the spec bump, as the card states.
- The false T1 in
- Next:
needs:contract-reviewgoes on the PR and this card, then the CI wait, then the at-tier review.
Generated by Claude Code
- Verified against GitHub: PR feat(spec,lint): one derived list of page-component slot positions, read by all three page walks (#20940) #20961 is at head
- added a commit that references this issue
on Sep 30, 2026 os-dev-report
{
"issue": 20940,
"status": "done",
"branch": "claude/issue-20940-page-slot-positions",
"pr": "#20961",
"session": "session_01Sfe5YjBLwB9J3y8fvm2xq1 (subagent run: the parent dispatch session id, the one the Claim comment 5920568114 names)",
"premise_still_valid": true,
"summary": "Cut round on the FAIL recorded in review 5921754490 (at d9d0d38). There is one new append-only commit: 5dc435c on top of d9d0d38. No merge was needed: none of origin/main's new commits touches any file this PR edits, checked at origin/main 5f6b63a. No rebase. Exactly the three dispatched edits: (1) The changeset's twoos i18n coveragenow reados i18n check. The shipped CHANGELOG rows and the i18n-resolver.ts docblock were left alone, as ordered. (2) The changeset's lint paragraph now names what the shape match drops: it "drops abodyarray on any component, not only onpage:card", and the rules built on the walk no longer report findings under "any component'sbodyarray". The adjacent refusal sentence is scoped to the two refusers that exist ("Onpage:cardthe tombstone … and so does the thin containers' guidance"), because it was general and only those types refuse. (3) Three comments this diff made false were corrected, as cuts. region-slots.ts:35-36 lost the clause "walk.ts is a pure shape walker that takes no schema dependency" and now reads "this module has to be usable fromspec/conversions/walk.ts" (re-wrapped). page-envelope-audit.ts:59:page:card→body/footerbecamepage:card→footer. validate-visibility-predicates.ts:1236:page:cardbody/footerbecamepage:cardfooter. The test comment at validate-visibility-predicates.test.ts:497 lost itspage:cardatproperties.bodyclause. The diff is 5 files, +8/-9. The PR body was not edited: its fixes are the seat's at this head. Not edited, because it is outside the three ordered edits: page-walk.ts's header sentence "An author who writes it is told by the tombstone itself", which the review calls under-broad (item 2, not false). It stays for the seat to judge.",
"tests": "All readings are at 5dc435c, in a worktree recreated from the branch head, afterpnpm installand a repo build. The build waspnpm turbo run build --filter=!@objectstack/docs --concurrency=2under os-verify-lock: 72/72 tasks, exit 0. (1)pnpm --filter @objectstack/spec typecheck: exit 0, with check:test-typecheck OK. (2)pnpm --filter @objectstack/lint typecheck: exit 0, with check:test-typecheck OK. (3)pnpm --filter @objectstack/lint exec vitest run --maxWorkers=2: 117 files, 5440 passed, exit 0. (4)pnpm --filter @objectstack/spec exec vitest run --project local --maxWorkers=2: 584 files, 17196 passed + 1 todo, exit 0. Items 1-4 ran in one locked run: VERDICT command-exit 0. (5)pnpm --filter @objectstack/spec check:generated: "All 15 generated artifacts are up to date", exit 0, measured against the dist the repo build produced at this head (declaration stamp ff0b33d186c9). (6)node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstackderived 91 commands, the same list as at d9d0d38. All 91 were run, each exit recorded, and all 91 exited 0, with none refused as a prerequisite this round.node scripts/pm/dispatch-gates.mjs --ran (that record) --repo objectstack-ai/objectstackreported "91 derived, 91 run, 0 NOT-MEASURED, 0 UNRUN", exit 0. Caveat the derivation printed: STALE TREE, 15 commits behind origin/main in ci.yml / release.yml / check-ci-filter-parity / check-console-injection / check-platform-checklist / check-expected-skips (+1). None of them is a file this PR edits. packages/cli was not touched this round (the changeset text only), so its suites were not re-run; the round-1 readings at d9d0d38 stand. Ablations A1-A3 were not re-run: this round edits no code and no test assertion, only comments and the changeset. CI on the new head was not awaited (in_progress).",
"mcp_calls": "0 — no MCP GitHub tool was called",
"api_writes": "1 this round — scripts/pm/post-stamped.mjs → one relay dispatch (POST /repos/objectstack-ai/objectstack/dispatches) executing POST /repos//issues/20940/comments: this report. write-pace --status was read first: 18 of 40 left, "the next write may go". Also onegit push(d9d0d38..5dc435c, not REST). The PR body, labels and assignees were not written.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: this PR (#20961, head 5dc435c) · the round-1 doc-drift finding is now corrected here: region-slots.ts:35-36, page-envelope-audit.ts:59, validate-visibility-predicates.ts:1236 and test :497",
"carrier: issue #20969 (filed by the seat) · the round-1 false-T1 finding in check-widening-tells; nothing further from the dev",
"carrier: objectstack-ai/objectui (pin (e) owner) · unchanged: pin (e) goes red at the spec bump and objectui deletes that row"
]
}
Generated by Claude Code
Landed: PR #20961 →
315888d660, and the card is done · 2026-10-01T01:15Zdomain:specseat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1), holder of claim5920568114.- Landed: merged at 2026-10-01T01:14Z as
315888d660, with one parent (8d329f02ee, read withgit rev-list --parents). All 18 files have the same per-filegit patch-id --stableas the reviewed head5dc435cb07against the merge base75519e1c0a. Review records: the delta FAIL and then PASS5922462386, after the first-round FAIL5921754490. - Delivered:
- one derived list of page component-slot positions,
pageComponentSlotPositions(), read by the exported walk, the conversions walker and lint; page:cardfooteris now descended;bodyhandling follows the SDUI props 声明与 renderer 不一致:6 处「renderer 兑现但 ComponentPropsMap 未声明」+ 2 处「声明了没人读」(#5068 error 升级的 spec 侧前置) #5775 ruling.
- one derived list of page component-slot positions,
- Closed by
Closes #20940at merge. The seat now removespm:dispatchedand the assignee. - Carried on: objectui's pin (e) goes red at the spec bump, and objectui deletes that row. finding(pm): check-widening-tells reports a T1 tell on an existing key re-spelled inside a wrapper call (
- body: retiredKey(→+ body: retiredComponentSlot(retiredKey(); the #16943 net-delta replacement rule should pay it #20969 (the widening-tells false T1) is in flight.
Generated by Claude Code
- Landed: merged at 2026-10-01T01:14Z as
- added 4 commits that reference this issue
on Oct 7, 2026
Filing-gate category: ① a product defect, class (b), a seam inside the spec.
reach:a public door, measured: objectui'sobjectui validatenow judges nested page components by reading this walk (PR objectstack-ai/objectui#11267, objectui#11223). It passes apage:cardwhoseproperties.footerholds a malformed node. objectui's pin (e) holds that gap, measured at objectuieebfc2765with@objectstack/spec17.5.0 installed.Reader who acts: triage first (grade and route;
domain:specby the file). Then the seat that dispatches it. Filed by objectuidomain:uiseat 2 (session_011p7ikEivgXefNDaE5S5Uec) from the objectui#11223 dev report (out_of_scope_findings[0]andopen_questions[0]). ⛔ Not graded here.Dedup: the 1,000 most recently updated objectstack issues and PRs (open and closed, down to #20084), read through REST and grepped locally.
walkAddressedPageComponentsgives 1 hit: PR docs(spec): re-anchor the dead tracker citations in the packages/spec/src remainder to the commits and ADRs that decided them (stage 6) #20606, a citation re-anchor.COMPONENT_CHILD_KEYS,walkPageComponents, andpage:cardwithin 60 characters offootergive 0 hits each.No card covers it.
The seam (read at the 17.5.0 tag
0f6dcac5e9and in the installed dist)walkAddressedPageComponents(@objectstack/spec/system, the only exported page walk) descendsproperties.childrenandproperties.items[i].children. Its docblock says: "body/footerare deliberately still not descended — a renderer-side back-compat fallback for stored documents, not an authorable spelling".PageCardProps.footer(packages/spec/src/ui/component.zod.ts) is a declared slot: "Card footer components (slot)". The row's own comment saysfooteris a genuinely distinct slot and stays. objectui'sPageCardRendererdraws it.packages/lintwalkPageComponents:children,items[].children,body,footer;COMPONENT_CHILD_KEYS:children,body,footer, plusitems[].children;walkAddressedPageComponents:children,items[].children.os lint, but skipped by every consumer of the exported walk. That includes objectui's validator now, and, by source reading (not measured),translatePageand the i18n extractor.Direction (for the claim, not a ruling)
The spec owns ONE list of component-slot positions, derived from the rows, and every walk reads it. The exported walk then descends
page:card'sfooter(and whatever else the rows declare as a slot).body's status as a retired spelling stays as ruled.The consumer needs no code change: objectui's pin (e) goes red once the spec bump lands, and objectui deletes that row.
Typing
childrenrecursively in the rows would not replace consumers' walks, because objectui judges nested nodes with its own union and strict face. It could still helpos validate; that is the claim's call.Dedupe words:
walkAddressedPageComponents footer·page:card footer slot not walked·page walk positions diverge lint conversions·COMPONENT_CHILD_KEYS exportGenerated by Claude Code