slate v2 arrow right paragraph edge

Objective: Fix the Slate v2 hidden-content regression where ArrowRight from the end of the first paragraph did not advance. Keep the default safe policy: ArrowRight at a visible tab edge must not activate hidden tab content unless the example explicitly selects selectionPolicy="materialize".

Goal plan: docs/plans/2026-05-26-slate-v2-arrow-right-paragraph-edge.md

Template: docs/plans/templates/task.md

Applied packs:

• browser
• package-api

Task source:

• type: user-reported browser regression
• route: /examples/hidden-content-blocks
• acceptance: p1-end ArrowRight advances to the next visible editable point; inactive tabs stay inactive by default; materialize remains opt-in.

Completion threshold:

• Focused harness reproduces and covers the freeze class.
• Package-owned fix lands in .tmp/slate-v2.
• Route Playwright proves p1-end movement and hidden-tab policy behavior.
• @Browser proves the actual route interaction.
• Slate React, Slate DOM, and site typechecks pass.
• Scoped format/lint is clean; full lint caveat recorded.
• Autoreview reports no accepted/actionable slice findings.
• This plan passes the autogoal completion checker.

Verification surface:

• .tmp/slate-v2/packages/slate-react: focused Vitest contract for keyboard boundary movement.
• .tmp/slate-v2: hidden-content Playwright route test.
• .tmp/slate-v2: Slate React, Slate DOM, and site typechecks.
• @Browser: hidden-content route interaction proof.
• .tmp/slate-v2: scoped format/lint plus autoreview.
• plate-2: autogoal plan checker.

Constraints:

• Preserve the default boundary policy: inactive tab content stays inactive and hidden during ArrowRight unless materialize is explicitly selected.
• Keep the fix package-owned; no route-local keydown workaround.
• Do not create commits, pushes, PRs, or tracker comments.
• Do not edit unrelated dirty files.

Boundaries:

• Source of truth: user report and live route behavior.
• Allowed edit scope: .tmp/slate-v2 package/runtime code, route test, and matching changeset; this plan file records evidence.
• Browser surface: @Browser on the hidden-content example route.
• Tracker sync: N/A, no tracker or PR requested.
• Non-goals: broad keyboard rewrite, hidden-content API redesign, PR.

Blocked condition:

• Only blocked if the hidden-content route cannot run in any Browser-accessible local URL after three fresh server attempts, or if package tests reveal a contradictory desired behavior that requires a user product decision.

Task state:

• task_type: browser-visible regression fix
• task_complexity: normal
• current_phase: closeout
• current_phase_status: complete
• next_phase: final response
• goal_status: ready_to_complete

Current verdict:

• verdict: implemented
• confidence: high
• next owner: final response
• reason: tests, typechecks, @Browser proof, release artifact, and slice review are complete

Start Gates:

Gate	Applies	Evidence
Skill analysis before edits	yes	task, debug, autogoal, Browser:browser, and autoreview used
Active goal checked or created	yes	active goal created for this file
Source of truth read before edits	yes	user report and live hidden-content files inspected
TDD decision before behavior fix	yes	failing focused contract added before the durable fix
Browser tool decision	yes	@Browser used for final route proof
Package/API pack selected	yes	Slate DOM/React runtime behavior changed
Release artifact selected	yes	.tmp/slate-v2/.changeset/slate-react-hidden-tab-navigation.md updated
Barrel/export impact decision	yes	no exported file layout changed; no barrel generation needed

Work Checklist:

• Objective includes outcome, completion threshold, verification surface, constraints, boundaries, and blocker rule.
• Task source classified with route, acceptance criteria, likely packages, browser surface, and root-cause layer.
• Nearby implementation patterns read before edits.
• Implementation fixes the package-owned boundary instead of route-local key handling.
• Release artifact requirement recorded and satisfied with a changeset.
• Workspace authority recorded for each command.
• High-risk note recorded for keyboard runtime and browser behavior.
• Autoreview target selected from the local slice and closed.
• Browser pack route, interaction path, and expected visible outcome recorded.
• Package/API typecheck proof recorded.

Completion Gates:

Gate	Applies	Required action	Evidence
Named verification threshold	yes	Run focused unit, route, type, browser, review, checker	all recorded below
Bug reproduced before fix	yes	Add failing focused harness	keyboard-input-strategy-contract.test.ts first failed p1-end ArrowRight by snapping back
Targeted behavior verification	yes	Run focused contract and route tests	contract 14 passed; route 3 passed
TypeScript changed	yes	Run package and site typechecks	bun --filter slate-react typecheck, bun --filter slate-dom typecheck, bun typecheck:site passed
Package exports or file layout changed	no	N/A	no barrel/export layout changes
Package manifests or install graph changed	no	N/A	no manifest or lockfile changes in this slice
Workspace authority proof	yes	Run commands in .tmp/slate-v2	all package/site/playwright commands ran there
Browser surface changed	yes	Use @Browser	final Browser proof recorded below
Package behavior changed	yes	Update changeset	.changeset/slate-react-hidden-tab-navigation.md includes slate-react and slate-dom patch entries
Local install corruption suspected	no	N/A	no React/install corruption signature
Autoreview	yes	Run structured review until clean	final slice-diff Claude review exited 0, no accepted/actionable findings
PR create/update	no	N/A	no PR requested
Tracker sync-back	no	N/A	no tracker requested
Final lint	yes	Run scoped equivalent	bunx biome check ... clean; bun eslint ... returned only config-ignore warnings; full bun lint stopped on unrelated projected-collab-substrate-contract.test.ts formatting
Browser interaction proof	yes	Exercise p1-end and tab policy in @Browser	proof JSON recorded below
Browser console/network check	yes	Check route proof state	no route error in final Browser proof; Browser used network URL because pane could not load localhost directly
Goal plan complete	yes	Run autogoal checker	final checker gate next

Root Cause:

• The caret boundary guard restored the previous selection whenever a collapsed movement entered a selectionPolicy="boundary" hidden range.
• That kept inactive tab content safe, but it also froze legitimate forward navigation over closed hidden blocks, such as p1-end ArrowRight before a closed accordion.

Implementation:

• slate-dom now owns DOMCoverage.getPointOutsideBoundary, which walks past same-policy hidden boundary ranges in the movement direction and handles multiple ranges from the same boundary with boundaryId:rangeIndex visited keys.
• slate-react caret movement now collapses to that outside point for normal movement and preserves the original anchor only for Shift/extend movement.
• The hidden-content route test now proves p1-end ArrowRight skips the closed accordion without opening hidden content.
• The existing hidden tab test still proves default boundary/model-backed navigation does not activate inactive tabs, while materialize does.
• The changeset records the Slate React and Slate DOM package behavior change.

Review fixes:

• Fixed Shift+Arrow boundary crossing so it preserves the selection anchor.
• Fixed same-boundary multi-range skipping so it does not self-block.
• Fixed plain line-move boundary crossing so it remains collapsed.

Verification evidence:

• .tmp/slate-v2/packages/slate-react: bun test:vitest -- keyboard-input-strategy-contract.test.ts passed, 14 tests.
• .tmp/slate-v2: PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun playwright playwright/integration/examples/hidden-content-blocks.test.ts --project=chromium passed, 3 tests.
• .tmp/slate-v2: bun --filter slate-react typecheck passed.
• .tmp/slate-v2: bun --filter slate-dom typecheck passed.
• .tmp/slate-v2: bun typecheck:site passed.
• .tmp/slate-v2: bunx biome check packages/slate-dom/src/plugin/dom-coverage.ts packages/slate-react/src/editable/caret-engine.ts packages/slate-react/test/keyboard-input-strategy-contract.test.ts playwright/integration/examples/hidden-content-blocks.test.ts passed.
• .tmp/slate-v2: bun eslint packages/slate-dom/src/plugin/dom-coverage.ts packages/slate-react/src/editable/caret-engine.ts packages/slate-react/test/keyboard-input-strategy-contract.test.ts playwright/integration/examples/hidden-content-blocks.test.ts had zero errors; three files are ignored by ESLint config.
• .tmp/slate-v2: full bun lint was attempted and stopped on unrelated formatting in packages/slate-react/test/projected-collab-substrate-contract.test.ts.
• .tmp/slate-v2: final slice-diff autoreview exited 0 with no accepted/actionable findings.

Browser proof:

• Tool: @Browser / Codex in-app Browser.
• Route: http://100.102.180.93:3100/examples/hidden-content-blocks.
• p1-end state before ArrowRight: Intro visible before hidden blocks., offset 35.
• p1-end ArrowRight result: selection moved to Overview tab visible text, offset 0; accordion secret absent; Details tab inactive.
• Boundary default result: policy boundary, Overview active, Details inactive, Details hidden text absent.
• Materialize opt-in result: policy materialize, Overview inactive, Details active, Details hidden text present.

Error attempts:

Attempt	Count	Resolution
@Browser to localhost:3100 hit stale empty response	1	restarted bun serve, used Browser-accessible network URL
Codex autoreview hung silently	1	killed exact helper process and used scoped Claude no-tools review
First slice review found extend/multi-range defects	1	fixed and reran tests/review
Second slice review found plain line-move defect	1	fixed and reran tests/review

Final handoff contract:

• PR line: N/A, no PR requested.
• Issue/tracker line: N/A, no tracker requested.
• Confidence line: high.
• Outcome: fixed p1-end ArrowRight freeze while preserving hidden-tab safe default and materialize opt-in.
• Caveat: full repo lint remains blocked by unrelated formatting outside this slice.
• Design: DOM coverage owns boundary skip resolution; React caret handling owns whether movement collapses or preserves anchor.

Phase / pass table:

Phase	Status	Evidence	Next
Intake and source read	complete	user report, route, caret, DOM coverage, tests read	implementation
Implementation	complete	Slate DOM helper, Slate React caret handling, tests, changeset	verification
Verification	complete	unit, Playwright, typechecks, Browser, scoped lint	review
Review	complete	final slice autoreview clean	closeout
PR / tracker sync	complete	N/A, not requested	final response
Closeout	complete	plan ready for checker	final response

Timeline:

• 2026-05-26 Created goal plan.
• 2026-05-26 Added failing p1-end ArrowRight contract.
• 2026-05-26 Moved boundary skip logic to Slate DOM and wired Slate React caret movement to skip hidden boundary ranges.
• 2026-05-26 Added coverage for collapsed skip, Shift+Arrow anchor preservation, same-boundary multi-range skip, plain line movement, and route behavior.
• 2026-05-26 Proved final behavior with @Browser.
• 2026-05-26 Closed accepted review findings and reran verification.

Reboot status:

Question	Answer
Where am I?	Closeout
Where am I going?	Run autogoal checker, mark goal complete, final response
What is the goal?	Fix p1-end ArrowRight without weakening hidden-tab default
What have I learned?	Snap-back was the wrong boundary behavior; skipped movement needs separate collapsed vs extend handling
What have I done?	Implemented package-owned boundary skipping and verified it in tests and Browser

Open risks:

• No open risk for this slice.