Slate v2 Selection / Clipboard Follow-up Clusters Ralplan

Current Verdict

Do not rearchitect the editor runtime for this lane.

Cluster 19 has already had the important scroll-selection work. This plan must not spend another pass on the same architecture unless a fresh regression test fails. Keep #5088 and #5473 as related unless Ralph adds exact after-update/delete-empty-paragraph browser proof.

The next useful execution work is:

cluster 20: add exact browser proof for blur and unfocused selection updates;
cluster 22: add only honest, testable WebKit/native-input guards; do not claim real Safari spellcheck correctness from Playwright;
cluster 25: preserve existing inline-boundary proof and decide whether #3150 deserves a narrower exact browser row or stays related to #3148;
cluster 21: preserve existing inline-void clipboard proof and keep #4802 as Improves, not Fixes, unless a named external-editor repro is proven.

Initial score: 0.88. Current score after related-issue-discovery pass: 0.89. Current score after issue-ledger pass: 0.90. Current score after intent-boundary-decision-brief pass: 0.91.

Intent And Boundary

Intent: process the next non-Android selection/input/clipboard issue clusters without reopening solved architecture or overclaiming browser-specific behavior.

Desired outcome: a Ralph-ready execution plan with focused tests, conservative issue claims, and no public API churn unless exact proof exposes a real gap.

In scope:

Slate React selection import/export when blur or unfocused updates happen.
WebKit/macOS-testable Cyrillic/native-input guard rows.
Inline DOM-end selection and inline boundary typing/paste rows.
Inline void browser clipboard payload, paste, cut, and external target proof.
Issue ledger sync when execution changes a claim.

Non-goals:

Android or iOS device claims.
Real Safari spellcheck underline/dictionary assertions from Playwright.
Rewriting the scroll-selection architecture from cluster 19.
New Plate product APIs.
Closing #4802 against unnamed third-party editors.

Decision boundaries:

Ralph may add focused unit/browser tests and small runtime fixes in .tmp/slate-v2.
Ralph may promote #4376, #5171, or #3150 only with exact browser proof.
Ralph must keep #5095 and #5096 related unless real Safari/device proof exists. WebKit smoke tests can only be guardrails.
Ralph must preserve existing fixed claims for #3148 and #4806.

Unresolved user-decision points: none. "22: go only what is testable" means desktop/WebKit guardrails are allowed, device/spellcheck claims are not.

Intent-boundary pass result:

intent, desired outcome, in-scope behavior, non-goals, and decision boundaries are explicit;
no user question is needed;
the weakest boundary is cluster 22, and it is resolved by the explicit rule: WebKit/native-input guardrails are allowed, but real Safari spellcheck claims require real Safari/device evidence;
cluster 19 is explicitly outside this lane unless a fresh test fails.

Decision Brief

Principles:

Browser-specific claims need matching browser proof.
Native spellcheck is not equivalent to typed-text stability.
Current public APIs stay unless a repro proves they are insufficient.
Existing exact fixes are regression gates before new work.

Top drivers:

Cluster 20 still has related rows only for Safari blur and Firefox unfocused update behavior.
Cluster 22 is mostly not automatable without real Safari spellcheck evidence.
Cluster 25 and 21 already have strong adjacent exact fixes.

Options:

Option	Pros	Cons	Verdict
Combined follow-up lane	Reuses selection and clipboard setup; keeps issue accounting together	Mixed proof types require strict claim discipline	Chosen
Separate plan per cluster	Cleaner paper trail per issue	Too much overhead for mostly proof-preservation work	Rejected
Full runtime rewrite	Could unify selection/input/clipboard under one abstraction	Wasteful; current owners are already correct	Rejected

Chosen option: one execution lane with independent proof rows and conservative claim rules.

Consequences:

This plan may finish with no new Fixes claim.
Cluster 22 probably stays related after execution.
Cluster 19 remains out of scope unless proof regresses.

Follow-ups:

If cluster 20 exposes a missing runtime abstraction, split a dedicated selection-frame plan.
If #4802 needs closure, define a named external-editor fixture instead of pretending a generic contenteditable target is enough.

Decision-brief pass result:

principles, top drivers, viable options, rejected alternatives, consequences, and follow-ups are recorded;
the chosen option remains one combined execution lane because the clusters share runtime proof setup but not claim semantics;
the rejected full-runtime rewrite is intentionally kept rejected because live source already has clear owners for selection, native input, and clipboard;
the plan is ready for the research/source refresh pass.

Source-Backed Architecture North Star

Keep one runtime owner per behavior:

selection runtime owns focus, blur, DOM import/export, and stale-selection prevention;
native input runtime owns beforeinput/input policy and Safari/WebKit guard behavior;
DOM clipboard runtime owns DataTransfer formats and Slate fragment transport;
browser examples prove user-visible editing, copy, paste, cut, and follow-up typing.

No public API change is accepted in the current-state pass.

Current Source Read

Cluster and issue ledgers:

docs/slate-issues/gitcrawl-recluster-map.md maps cluster 20 to #4376 and #5171, cluster 22 to #5095 and #5096, cluster 25 to #3148/#3150, and cluster 21 to #4802/#4806.
docs/slate-v2/ledgers/issue-coverage-matrix.md keeps #4376, #5171, #5095, and #5096 related.
docs/slate-v2/ledgers/issue-coverage-matrix.md already claims Fixes #3148 and Fixes #4806.
docs/slate-v2/ledgers/issue-coverage-matrix.md keeps #3150 related and #4802 improved.

Live Slate v2 source:

.tmp/slate-v2/packages/slate-react/src/editable/selection-reconciler.ts:226 owns blur/focus selection policy; WebKit blur currently removes DOM ranges.
.tmp/slate-v2/packages/slate-react/src/editable/runtime-selection-engine.ts:30 owns throttled selectionchange import and model/native ownership traces.
.tmp/slate-v2/packages/slate-react/src/editable/selection-runtime.ts:33 owns selection export policy after commits, including deferred DOM export for content-changing commits.
.tmp/slate-v2/packages/slate-react/test/selection-runtime-contract.test.ts:42 through 340 covers model-vs-DOM export policy, content-changing commit deferral, and text-input repair skip behavior.
.tmp/slate-v2/packages/slate-react/src/components/editable.tsx:86 exposes scrollSelectionIntoView, onDOMBeforeInput, and native div props; :322 passes spellCheck through when beforeinput support exists.
.tmp/slate-v2/packages/slate-react/src/editable/model-input-strategy.ts:69 owns React onInput handling and native input repair.
.tmp/slate-v2/packages/slate-react/src/editable/runtime-before-input-events.ts:92 owns DOM beforeinput routing and model/native input decision setup.
.tmp/slate-v2/playwright/integration/examples/inlines.test.ts:130 proves following text start stays distinct from inline end; :305 proves padded inline caret placement.
.tmp/slate-v2/playwright/integration/examples/mentions.test.ts:249 proves deterministic selected mention clipboard payload; :328 proves selected mention cut and model-owned caret repair.
.tmp/slate-v2/playwright/integration/examples/document-state.test.ts:256 proves a document-level spellcheck state field drives the spellCheck prop, but it does not prove native Safari spellcheck.

Already reworked cluster 19:

docs/plans/2026-05-11-slate-v2-scroll-selection-visibility-ralplan.md accepts selection import first, scroll request second, post-update visibility third.
.tmp/slate-v2/packages/slate-react/test/app-owned-customization.tsx:501 proves scroll forwarding.
.tmp/slate-v2/packages/slate-react/test/app-owned-customization.tsx:532 proves remote selection updates skip scroll forwarding.

Ecosystem Strategy Synthesis

System	Source	Mechanism	Avoids	Steal	Reject	Slate target	Verdict
ProseMirror	`docs/research/sources/editor-architecture/scroll-selection-visibility-runtime.md`	Transactions separate selection, update, and scroll intent	stale scroll/caret reveal	post-update selection and scroll proof discipline	integer positions as public model	cluster 20 browser rows should prove post-update selection ownership	partial
Lexical	`docs/research/decisions/slate-v2-data-model-first-react-perfect-runtime.md`	Native events feed an update lifecycle with explicit selection reconciliation	browser timing patches scattered in React handlers	model/native ownership traces and DOM import before native input	class-node/$ helper API	keep Slate data-model first, runtime-owned selection/input lanes	agree
Tiptap	`docs/research/sources/editor-architecture/scroll-selection-visibility-runtime.md`	Product command delegates focus/scroll to ProseMirror engine	app commands owning engine truth	simple app-facing boundary only	product wrapper APIs in raw Slate	keep raw Slate unopinionated; Plate can wrap later	partial
React 19.2	`docs/research/sources/editor-architecture/react-19-2-external-store-and-background-ui.md`	External-store subscriptions and background UI primitives	broad render churn during editor interactions	selector-first runtime proof and urgent visible editing	React as editor engine	keep React as projection consumer, not selection truth	agree
Slate v2 clipboard	`docs/plans/2026-05-23-slate-v2-clipboard-fragment-serialization-ralplan.md`	Core model fragment, DOM DataTransfer, React event routing split	product parser leaking into raw Slate	preserve exact inline-void clipboard proof	public raw `editor.clipboard` namespace	cluster 21 is proof preservation plus possible named external fixture	agree

Public API Target

Keep:

Editable scrollSelectionIntoView?: (editor, domRange) => void
Editable onDOMBeforeInput
normal React div props such as spellCheck
DOM-hosted editor.api.clipboard capability

Do not add:

public blur-selection policy props;
public Safari spellcheck workaround props;
a new raw Slate clipboard namespace;
a broad inline-boundary normalization hook for #3150.

Internal Runtime Target

Cluster 20:

prove blur clears focus without losing model selection identity;
prove unfocused editor changes do not import invalid browser selections;
keep selectionchange ownership explicit in kernel traces when possible.

Cluster 22:

preserve native spellCheck pass-through;
prove Cyrillic text insertion survives WebKit/native input routing when spellCheck is enabled;
do not assert browser spellcheck dictionary/UI from Playwright.

Cluster 25:

preserve inline end vs following text start distinction;
only add a #3150 exact row if the original inline DOM end repro can be expressed as a modern browser contract.

Cluster 21:

preserve selected inline void clipboard payload, paste, cut, and caret repair;
keep #4802 improved unless execution uses a named external editor target.

Hook / Component / Render DX Target

No app example should need dirty focus restoration helpers or custom DOM selection imports for these clusters.

Expected app DX:

tsx

<Editable spellCheck scrollSelectionIntoView={onScrollSelectionIntoView} />

Clipboard and inline/void behavior should be runtime behavior, not example app logic.

Plate Migration Backbone

Plate can consume these outcomes without special adapters:

blur/focus selection stability becomes a raw Slate guarantee;
spellcheck remains a normal DOM prop plus document-state/user preference;
inline-boundary behavior stays in Slate runtime, not Plate plugins;
product-specific paste/export policy remains a Plate extension lane.

No current-version Plate adapter is required for this plan.

slate-yjs Migration Backbone

Collaboration pressure is limited to selection identity and deterministic clipboard operations:

remote/unfocused changes must not import the wrong local DOM selection;
clipboard cut/paste must produce deterministic operations;
no yjs-specific API is needed.

Research / Source Refresh Pass

Status: complete.

Compiled research checked:

docs/research/sources/editor-architecture/scroll-selection-visibility-runtime.md confirms the cluster 19 direction: selection import first, update second, post-update scroll/visibility third. No cluster 19 rewrite belongs in this lane.
docs/research/sources/editor-architecture/react-19-2-external-store-and-background-ui.md confirms React should stay the projection/subscription layer. It should not become selection truth.
docs/research/decisions/slate-v2-data-model-first-react-perfect-runtime.md confirms the current target: Slate data model first, transaction/runtime-owned selection and native-input lanes, React as consumer.
docs/research/decisions/clipboard-and-delete-commands-need-explicit-lanes.md confirms clipboard and delete behavior should stay explicit lanes, not incidental editing side effects.

Live source checked:

.tmp/slate-v2/packages/slate-react/src/editable/selection-reconciler.ts:227 through 305 keeps blur/focus selection policy local to the reconciler.
.tmp/slate-v2/packages/slate-react/src/editable/runtime-selection-engine.ts:47 through 116 keeps selectionchange import and ownership traces centralized.
.tmp/slate-v2/packages/slate-react/src/editable/selection-runtime.ts:33 through 240 keeps model-to-DOM export policy explicit, including deferred DOM export after content-changing commits.
.tmp/slate-v2/packages/slate-react/src/components/editable.tsx:86 through 104 keeps scrollSelectionIntoView public and narrow; :322 through 326 keeps spellCheck as a normal DOM prop pass-through when supported.
.tmp/slate-v2/packages/slate-react/src/editable/model-input-strategy.ts:72 through 188 keeps native text-input repair in the input lane.
.tmp/slate-v2/packages/slate-react/test/selection-runtime-contract.test.ts:42 through 163 and :220 through 340 already cover the selection export contract, deferred export, and repaired text-input skip behavior.
.tmp/slate-v2/playwright/integration/examples/inlines.test.ts:130 through 162 and :305 through 338 already cover inline-boundary proof needed for cluster 25 preservation.
.tmp/slate-v2/playwright/integration/examples/mentions.test.ts:249 through 400 already covers selected inline void copy/cut payload and caret repair needed for cluster 21 preservation.
.tmp/slate-v2/playwright/integration/examples/document-state.test.ts:256 through 286 proves a document state field can drive spellcheck state, but does not prove native Safari dictionary/UI behavior.

Decision:

No target change.
Cluster 20 still needs exact browser proof before fixed claims.
Cluster 22 remains guardrail-only unless real Safari/device proof exists.
Clusters 25 and 21 should preserve existing fixed/improved rows rather than widen claims from generic coverage.

Issue Ledger Accounting

ClawSweeper pass status: applied through existing completed ledgers. No broad live GitHub discovery is needed.

Related-issue discovery pass result:

direct live rows exist for every target issue in docs/slate-issues/gitcrawl-live-open-ledger.md;
manual sync rows already exist for #4376, #5171, #5095, #5096, #3148, #3150, #4802, and #4806;
fork dossier sections already exist for every target issue;
adjacent focus/scroll pressure to carry as related-only context: #5538, #4995, #3893, #5867, and #5559;
adjacent inline-boundary proof rows to preserve: #4074, #3429, and #4618;
adjacent void/delete proof rows to preserve while touching inline void clipboard behavior: #3991 and #4301;
no issue classification changed in this pass, so dossier, matrix, manual sync, and PR reference files remain unchanged.

Issue-ledger pass result:

docs/slate-issues/package-impact-matrix.md keeps this whole lane in the runtime boundary: selection/focus/DOM bridge, mobile/input semantics, and clipboard serialization belong to slate-react plus slate-dom, not raw slate core.
docs/slate-issues/requirements-from-issues.md reinforces the same boundary: DOM selection and input ownership must be explicit, and the runtime should stop leaking browser/input-method debt into normal editing.
docs/slate-issues/benchmark-candidate-map.md does not add a required benchmark owner for this lane. No performance benchmark is required unless execution changes hot clipboard or selection loops.
docs/slate-issues/test-candidate-map/4541-4392.md marks #4376 as ready-with-minor-setup with public seam "selection persistence across blur in Safari".
docs/slate-issues/test-candidate-map/5246-5130.md marks #5171 as ready-now with public seam "Firefox invalid selection update while blurred".
docs/slate-issues/test-candidate-map/5129-5066.md marks #5095 and #5096 as blocked-on-repro, so they stay related unless new real Safari evidence appears.
docs/slate-issues/test-candidate-map/3313-2733.md marks #3148 and #3150 as upstream-boundary constraints, not direct local red-test candidates. Existing exact #3148 browser proof remains enough; #3150 should not become a fixed row from generic coverage.
docs/slate-issues/test-candidate-map/4839-4742.md marks #4806 and #4802 as ready-now; execution should preserve #4806 and can only promote #4802 with a stronger external-editor fixture.
no contradiction was found between package ownership, requirements, candidate maps, current matrix rows, and the plan issue matrix.

Issue	Cluster	Claim	Why	Proof route	V2 sync ledger	PR line
`#5088`	19	Related	Scroll after-update family already reworked but exact after-update row is not claimed here.	preserve existing scroll tests; add only if regression appears	unchanged	related matrix only
`#5473`	19	Related	Delete-empty-paragraph scroll callback needs exact browser proof; not this lane unless regression emerges.	optional focused browser row	unchanged	related matrix only
`#4376`	20	Candidate Fixes	Safari blur null-selection may be provable through WebKit blur/focus browser row.	WebKit/Chromium selection blur browser proof plus package selection contract	update if exact proof passes	fixed line only if exact proof passes
`#5171`	20	Candidate Fixes	Firefox invalid unfocused update may be provable through a focused Firefox browser row.	Firefox browser row for unfocused content update plus selection contract	update if exact proof passes	fixed line only if exact proof passes
`#5095`	22	Related	iOS Safari spellcheck is `blocked-on-repro` and requires real iOS Safari proof.	no fixed claim without device proof	unchanged	related matrix only
`#5096`	22	Related	macOS Safari spellcheck is `blocked-on-repro`; WebKit typed-text stability is only a guardrail.	WebKit Cyrillic/native-input guard only	unchanged unless real Safari proof exists	related matrix only
`#3148`	25	Fixes	Existing browser proof distinguishes inline end and following text start.	preserve inlines Playwright row	unchanged	keep existing Fixes line
`#3150`	25	Related	Candidate map says this is upstream-boundary tracking, not a direct red-test candidate.	preserve `#3148` proof; no fixed claim	unchanged	related matrix only
`#4802`	21	Improves	External contenteditable target is proven; named other-editor behavior is not.	preserve mentions clipboard payload row; optional named fixture	unchanged	keep Improves
`#4806`	21	Fixes	Existing browser proof covers selected inline void copy, paste, cut, payload, and caret repair.	preserve mentions Playwright rows	unchanged	keep existing Fixes line
`#5538`	singleton adjacent	Related	Programmatic focus/select-to-end scroll behavior is adjacent scroll-selection pressure, not part of this execution lane.	no new proof unless cluster 19 regresses	unchanged	related matrix only
`#4995`	singleton adjacent	Related	Scroll customization pressure is cluster 19-adjacent and already covered by app-owned scroll callback proof.	preserve app-owned customization tests	unchanged	related matrix only
`#3893`	singleton adjacent	Related	External button focus is related focus-state pressure; cluster 20 may strengthen it but does not target exact closure.	no exact proof in this lane	unchanged	related matrix only
`#5867`	singleton adjacent	Related	Selected mention focus restoration overlaps inline void/focus behavior, but exact `DOMEditor.focus` closure is not claimed.	no exact proof in this lane	unchanged	related matrix only
`#5559`	singleton adjacent	Related	Shift-click through voids is related selection import pressure, but this lane targets blur/unfocused and inline clipboard.	no exact proof in this lane	unchanged	related matrix only
`#4074`, `#3429`	cluster 5 adjacent	Fixes	Existing inline-boundary rows must remain green when touching cluster 25.	preserve inlines Playwright rows	unchanged	keep existing Fixes lines
`#4618`	cluster 5 adjacent	Not claimed	The plan does not add a public `normalizePoint` hook.	no proof route	unchanged	related matrix only
`#3991`, `#4301`	cluster 17 adjacent	Fixes	Existing void delete/selection rows must remain green when touching inline void clipboard behavior.	preserve images Playwright rows if clipboard code changes	unchanged	keep existing Fixes lines

PR-description status: unchanged in this pass. Execution must update it only if new exact fixed/improved claims are added.

Fork issue dossier status: existing sections cover all touched issues. Execution must refresh only sections whose claim changes.

Revision / Issue Sync Pass

Status: complete.

Checked:

docs/slate-v2/ledgers/issue-coverage-matrix.md
docs/slate-v2/ledgers/fork-issue-dossier.md
docs/slate-issues/gitcrawl-v2-sync-ledger.md
docs/slate-v2/references/pr-description.md

Decision:

No ledger or PR-reference write is needed in this planning pass.
The plan hardened no-overclaim rules but did not promote, demote, or rewrite any current fixed/improved claim.
#3148 and #4806 remain fixed; #4802 remains improved.
#4376, #5171, #5095, #5096, and #3150 remain related/candidate rows until Ralph produces exact browser proof.
The PR description already contains fixed/improved rows for exact claims and does not list the candidate related issues as fixed.

Sync rule for Ralph:

If execution proves #4376, #5171, or #3150 exactly, update the coverage matrix, fork dossier, manual sync ledger, and PR reference in the same slice.
If execution only preserves existing behavior or adds guardrails, leave those rows related and update only the plan/execution evidence.

Legacy Regression Proof Matrix

Surface	Required proof	Current owner
Blur selection	browser row plus package selection runtime contract	`.tmp/slate-v2/packages/slate-react/test/selection-runtime-contract.test.ts`
Unfocused update	Firefox browser row or no fixed claim	new Ralph test
WebKit Cyrillic guard	WebKit typed text plus `spellCheck` pass-through	new Ralph test or document-state extension
Inline DOM end	existing inlines rows plus optional `#3150` exact row	`.tmp/slate-v2/playwright/integration/examples/inlines.test.ts`
Inline void clipboard	existing mentions rows	`.tmp/slate-v2/playwright/integration/examples/mentions.test.ts`

Browser Stress / Parity Strategy

Chromium: preserve existing inline and clipboard rows.
Firefox: target cluster 20 unfocused update behavior.
WebKit: target cluster 20 blur behavior and cluster 22 typed Cyrillic guard.
Mobile/iOS: no claim.

Applicable Implementation-Skill Review Matrix

Lens	Applicability	Finding	Plan delta
Vercel React	applied	Selection/input should stay external-store/runtime-owned; no broad render subscriptions.	Require package selection contract and focused browser rows.
performance-oracle	skipped	No new hot algorithm or large-doc path proposed.	No perf benchmark required.
performance	applied lightly	Clipboard/selection rows must not widen to degraded virtualization behavior.	Keep DOM/native behavior proof explicit.
tdd	applied	Execution should add one focused failing browser row before runtime changes.	Red-green browser proof required for any new claim.
shadcn	skipped	No UI chrome or component composition change.	None.
react-useeffect	skipped	No new effect design proposed.	None.

High-Risk Deliberate Mode

Triggered because the lane touches browser selection, focus, native input, and clipboard behavior.

Pre-mortem:

A Firefox/WebKit test passes synthetically but does not match the real issue.
A fix preserves model selection but breaks native text input/spellcheck.
A clipboard change fixes inline voids but leaks neighboring text or FEFF.

Proof plan:

unit/package selection runtime contract;
browser rows per issue family;
no fixed claims for non-assertable native spellcheck UI;
rerun existing inlines and mentions regression rows before claim sync.

Blast radius:

slate-react selection/input runtime;
slate-dom clipboard runtime if cluster 21 changes;
inlines, mentions, document-state examples;
issue coverage matrix, fork dossier, PR reference.

Rollback answer:

If exact browser proof is not possible, keep the issue related/improved. Do not bend public API around unproven behavior.

Pressure / objection pass result:

Keep the architecture. The correct owner is still the runtime lane, not app examples, Plate wrappers, or a new public API.
Strengthen claim discipline. #4376, #5171, #5095, #5096, #3150, and #4802 must not be promoted from adjacent coverage.
Require negative assertions where the failure mode is leak/collapse: blur must not destroy model selection identity, unfocused updates must not import stale DOM selection, inline text must not enter the link, and inline void clipboard payload must not include neighboring text or FEFF.
Keep cluster 22 as testable guardrail only. WebKit typed-text stability with spellCheck enabled is useful; it is not a Safari spellcheck UI fix.
Keep cluster 21 split: #4806 can remain fixed through selected inline void clipboard proof; #4802 remains improved unless a named external editor is proven.

High-risk execution rules for Ralph:

Risk	Required guard	Claim rule
WebKit/Firefox row passes but does not match the issue	Test the exact browser state transition, not just final text	keep candidate/related unless the original issue contract is represented
Blur fix clears DOM selection and loses Slate selection	Assert model selection identity after blur and after refocus/follow-up edit	`#4376` only with exact proof
Unfocused update imports stale browser selection	Assert commit/update does not mutate selection from inactive DOM	`#5171` only with Firefox-focused proof
Spellcheck guard becomes a fake Safari fix	Assert typed Cyrillic/native-input stability only	keep `#5095/#5096` related without real Safari proof
Inline-boundary row hides the `#3150` repro	Preserve existing inline tests and add an exact row only if expressible	keep `#3150` related otherwise
Clipboard test proves generic contenteditable only	Keep deterministic Slate fragment and external paste assertions	keep `#4802` improved unless named target proof exists

Pressure score: 0.93. Remaining risk is execution proof, not plan shape.

Hard Cuts And Rejected Alternatives

Cut reworking cluster 19 in this lane.
Cut fixed claims for real Safari spellcheck without real Safari/device proof.
Cut a public blur-selection API.
Cut raw Slate product clipboard serializers.
Cut claiming #4802 fixed from a generic contenteditable target.

Slate Maintainer Objection Ledger

Change	Likely objection	Answer	Evidence	Verdict
Add cluster 20 browser rows	"Are we testing browser quirks instead of Slate behavior?"	Yes, because these issues are browser-selection interop bugs; the runtime owns import/export timing.	`selection-reconciler.ts`, `runtime-selection-engine.ts`	keep
Add cluster 22 guard only	"Why not fix the spellcheck issue?"	Because real spellcheck UI/dictionary behavior is not assertable in Playwright WebKit. Guard native input without overclaiming.	`Editable` spellCheck pass-through and release-proof mobile gate	keep
Preserve `#4802` as Improves	"Why not close it if external paste works?"	The issue is about other editors; generic contenteditable is materially helpful but not exact.	`mentions.test.ts` clipboard rows	keep
Avoid public escape hatches	"Should users get props for blur policy, point normalization, or clipboard serialization?"	No. These are runtime correctness concerns or product extension concerns; raw Slate should stay small.	Public API target and rejected alternatives	keep
Keep cluster 19 out	"Why not revisit scroll selection since it is nearby?"	Because that lane is already reworked. Touch it only if the new focused tests regress.	scroll-selection visibility plan and app-owned customization proof	keep

Pass Schedule And Pass-State Ledger

Pass	Status	Evidence added	Plan delta	Open issues	Next owner
current-state-read	complete	live ledgers, current `.tmp/slate-v2` selection/input/clipboard owners, research pages	created this plan and scoped cluster 19 out	none	Slate Ralplan
related-issue-discovery	complete	live ledger rows, manual sync rows, fork dossier, coverage matrix, and PR reference searched for target plus adjacent issue refs	added adjacent related-only and preserve-only rows; no ledger files changed	none	Slate Ralplan
issue-ledger-pass	complete	package-impact matrix, requirements, benchmark map, and test-candidate maps checked	tightened blocked-on-repro and not-direct-test-candidate rows for clusters 22 and 25	none	Slate Ralplan
intent-boundary-decision-brief	complete	intent/boundary record and decision brief validated	added pass result notes and kept cluster 22 proof boundary explicit	none	Slate Ralplan
research-source-refresh	complete	compiled research and current `.tmp/slate-v2` source/test owners checked	no target change; recorded exact proof boundaries for clusters 20, 22, 25, and 21	none	Slate Ralplan
pressure-objection-high-risk	complete	high-risk matrix, negative assertion rules, and maintainer objections hardened	strengthened no-overclaim rules; no architecture change	none	Slate Ralplan
revision-and-issue-sync	complete	coverage matrix, fork dossier, manual sync ledger, and PR reference audited	no external ledger/reference write needed because no claim changed	none	Slate Ralplan
closure-final-gates	complete	all earlier passes complete; final handoff and gates written	closed Slate Ralplan planning lane; execution belongs to later Ralph	none	Ralph

Plan Deltas From Review

Added a new plan lane instead of extending the completed non-Android cluster plan.
Recorded cluster 19 as already reworked and out of scope.
Split cluster 22 into testable guardrails versus non-claim Safari spellcheck UI.
Preserved existing cluster 25 and 21 fixed/improved rows.
Named exact issue claim rules for Ralph execution.

Open Questions

None for planning. What would change the decision:

real Safari/macOS spellcheck automation exists;
user wants a named external-editor fixture for #4802;
current inlines/mentions/browser rows fail under Ralph.

Implementation Phases With Owners

Ralph execution status:

Status: done.
Current pass: verification-sweep-pass.
Current owner: .tmp/slate-v2 slate-react browser selection proof.
Result: cluster 20 exact blur/unfocused selection rows promote #4376 and #5171; cluster 22 stays guardrail-only; clusters 25 and 21 preserve existing #3148/#4806 proof while keeping #3150 related and #4802 improved.
Continuation prompt: active goal state.
Completion state: active goal state.

Phase 1, cluster 20:

complete;
added WebKit blur/refocus/follow-up typing proof for #4376;
added Firefox unfocused editor update proof for #5171;
no runtime fix was needed because the proof rows passed;
synced #4376/#5171 across coverage matrix, fork dossier, manual sync ledger, open issues ledger, and PR reference.

Phase 2, cluster 22:

complete;
added WebKit Cyrillic/native-input guard with spellCheck enabled;
kept #5095/#5096 related because this is not real Safari spellcheck UI or dictionary proof.

Phase 3, clusters 25 and 21:

complete;
reran existing inlines rows for #3148 preservation;
reran selected mention clipboard rows for #4806 preservation;
kept #3150 related and #4802 improved.

Ralph Execution Results

Changed proof:

.tmp/slate-v2/playwright/integration/examples/document-state.test.ts adds WebKit blur/refocus/follow-up typing proof, Firefox unfocused update proof, and WebKit Cyrillic typed-input guard with spellCheck enabled.
.tmp/slate-v2/playwright/integration/examples/decorations-async.test.ts guards selection.anchorNode before passing it to Range.setEnd, closing the root typecheck blocker exposed during verification.

Claim changes:

#4376: Related -> Fixes.
#5171: Related -> Fixes.
#5095/#5096: unchanged Related.
#3150: unchanged Related.
#4802: unchanged Improves.
#3148/#4806: existing Fixes claims preserved.

Synced files:

docs/slate-v2/ledgers/issue-coverage-matrix.md
docs/slate-v2/ledgers/fork-issue-dossier.md
docs/slate-issues/gitcrawl-v2-sync-ledger.md
docs/slate-issues/open-issues-ledger.md
docs/slate-v2/references/pr-description.md

Verification:

bun typecheck:root
bun lint:fix
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/document-state.test.ts --project=webkit -g "focus leaves|unfocused editor|Cyrillic"
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/document-state.test.ts --project=firefox -g "unfocused editor"
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/inlines.test.ts --project=chromium -g "start of following text|padded inline"
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/inlines.test.ts --project=webkit -g "start of following text"
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/mentions.test.ts --project=chromium -g "selected mention"

Fast Driver Gates

Planning gate, cwd /Users/zbeyens/git/plate-2:

bash

node tooling/scripts/completion-check.mjs

Execution gates, cwd /Users/zbeyens/git/slate-v2:

bash

bun test ./packages/slate-react/test/selection-runtime-contract.test.ts
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/inlines.test.ts --project=chromium -g "inline|padded"
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/mentions.test.ts --project=chromium -g "mention"
PLAYWRIGHT_BASE_URL=http://localhost:3100 PLAYWRIGHT_RETRIES=0 bun playwright playwright/integration/examples/document-state.test.ts --project=webkit -g "spellcheck|Cyrillic"
bun --filter slate-react typecheck
bun lint:fix

Ralph may tighten greps after adding exact test names.

Final User-Review Handoff Outline

Verdict:

This plan is ready for Ralph execution.
No Slate v2 source/test/example code was changed by Slate Ralplan.
Architecture stays unchanged: selection, native input, and clipboard remain separate runtime-owned lanes.
Public API stays unchanged: no blur-selection prop, no Safari spellcheck prop, no raw Slate clipboard serializer API, and no public inline point normalization hook.
Cluster 19 stays out because scroll-selection visibility was already reworked.

Execution handoff:

Cluster 20: add exact blur/unfocused selection browser rows for #4376 and #5171; fix only after a red row; promote claims only with exact proof.
Cluster 22: add only the testable WebKit/native-input Cyrillic guard with spellCheck; keep #5095/#5096 related without real Safari/device proof.
Cluster 25: preserve existing #3148 inline-boundary proof; keep #3150 related unless the original repro becomes an exact browser row.
Cluster 21: preserve #4806 selected inline void clipboard proof; keep #4802 improved unless a named external editor is proven.
After execution, sync issue coverage, fork dossier, manual sync ledger, and PR reference only for claims that actually change.

Final Completion Gates

Status: done.

Closed:

related-issue discovery pass closed;
issue-ledger pass closed;
intent/decision/research/pressure passes closed;
revision/issue sync pass closed;
issue coverage/fork dossier/PR reference audited with no planning-time claim change;
final handoff emitted.

Deferred to Ralph execution:

.tmp/slate-v2 focused gates for any new behavior claim;
issue coverage/fork dossier/PR reference updates if execution promotes a candidate issue.