feat(client): sprint 36 — MetaScreen pattern, character creation, bookmarks, protocol v23 #134
Assignees
jpmschweitzer (Jeroen Schweitzer)
Clear assignees
jpmschweitzer (Jeroen Schweitzer)
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: jpmschweitzer/settled-reach#134
Reference in New Issue
Block a user
Summary
Sprint 36 client work — closes Phase 3 Atlas (unified nav chain) and establishes Phase 4 character-creation foundations (MetaScreen pattern, 4-tab creation flow, bookmark system, protocol v23).
HudGroups.CharacterProfilesignal payload.bookmark_catalogdecode + bookmark action encoding. Server sends catalog on connect; client surfaces it in the Bookmark tab.PROTOCOL_VERSIONlockstep handshake — new decision. Implementation tracked in #868.after_test()that was freeing GdUnit4 internals (caused the full client test run to hang indefinitely on the second compositor test). Deleted tautologicaltest_protocol_version_is_NNassertions (× 2 suites). Remaining pre-existing failures ticketed as #869, #870, #871.Test plan
make test-clientcompletes end-to-end (1243 tests, 16 errors, 28 failures — all attributed to #869/#870/#871, none introduced by this branch)gdlint client/scripts/ client/ui/— zero warningsgodot --headless --path client --quit— zero new script/parse/export errorsmake game→ main menu → New Game → character creation tabs (reviewer)Tickets
Closes: #618 (character creation), #680 (bookmark picker)
Implements D-166 Phase 3 criterion 1 (#844), D-192 (via #868)
Filed from triage: #868, #869, #870, #871
Consolidates the character creation TabContainer from 8 flat tabs (Body / Head / Hair / Clothing / Accessories / Debug plus the two being-added Skills / Bookmark) into 4 top-level tabs per Araminta's revised spec: Bookmark, Appearance, Skills, Debug. The existing five appearance sub-tabs (Body, Head, Hair, Clothing, Accessories) now live inside the Appearance tab as a horizontal segmented sub-navigation using the existing `_make_slot_btn()` pattern — consistent with the Clothing/Accessories slot row vocabulary. Selected sub-section uses existing ITEM_SELECTED_BG / BORDER styling. Structural changes: - New APPEARANCE_SUB_NAMES const lists the five sub-sections. - Renamed _tab_search → _appearance_search, _tab_grids → _appearance_grids. Scope changed from "top-level tabs" to "Appearance sub-sections" but index 0..4 semantics preserved. - Added _appearance_active_idx, _appearance_sub_btns, _appearance_sub_sections state. - _ready() builds exactly 4 top-level tabs; tab builders invoked explicitly per index. - New _build_bookmark_tab / _build_skills_tab render TEXT_DIM placeholder labels ("Bookmark content lands in Workstream 6", etc.) — actual content in W6/W8. - _build_appearance_tab constructs the sub-nav strip and stacks all 5 sub-sections up front with visibility-toggle swap (_on_appearance_sub_selected). Comment explains the up-front build choice and the free-and-rebuild fallback if performance regresses. - Existing _build_body_tab / _head / _hair / _clothing / _accessories / _debug remain unchanged — they now receive Appearance sub-section Controls as their tab argument instead of top-level tabs. _make_tab_vbox anchors full-rect in both parent contexts, so layout is preserved. Verification: - gdlint clean - godot --headless --path client --quit — no SCRIPT ERROR - test_protocol 62/62, test_client_p3 24/24, test_ui_framework_sprint15 54/54, test_implant_nav_stack 52/52, test_implant_registry 42/42, test_implant_app_lifecycle 36/36 Workstream 6 (Bookmark tab content: card list + detail view + location picker per Araminta's spec) lands next. W7 (location picker as a sub-component of Bookmark tab) follows. W8 (Skills stub content) is last. Hoshe's parallel Task #21 (test hygiene triage) commits separately. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>Fills in the Bookmark tab stubbed in W5 with the full spec from Araminta: card list + detail view + start-button gating. Also changes the creation_confirmed signal to carry a CharacterProfile instead of bare CharacterVisualDescriptor, consolidating bookmark + location selection into one payload object. Bookmark tab (left pane, 35%): - ScrollContainer over VBoxContainer of card Buttons, one per entry in GameState.bookmark_catalog. Each card: title (PRIMARY_TEXT, font_header 15px) / subtitle (DIM_TEXT, font_small 10px, clipped) / career badge (ACCENT_ACTIVE, all-caps). Selected state uses existing ITEM_SELECTED_BG + ITEM_SELECTED_BORDER. custom_minimum_size Vector2(180, 64). Detail view (right pane, 65%): - ImplantPanel composed via add_component: - ImplantHeader (bookmark.title, bookmark.subtitle) - ImplantSeparator - ImplantTextBlock (flavor, autowrap, PRIMARY_TEXT) - ImplantSeparator - ImplantDataRow CAREER (accent_active) / CAPITAL (accent_positive, format "%d Tractus") / STARTING LOCATION - ImplantSeparator - [location picker space reserved — W7 fills it] Selection: - Card click stores _selected_bookmark_id, auto-assigns _selected_location_id from bookmark.default_location, rebuilds detail view. - Start button (footer) gated on both _selected_bookmark_id and _selected_location_id non-empty. - Randomize while Bookmark tab is active picks a random bookmark + one of its allowed_locations and skips appearance randomization. Signal contract change: - creation_confirmed(profile: CharacterProfile) replaces creation_confirmed(descriptor: CharacterVisualDescriptor). - CharacterProfile now extends RefCounted (was Resource) with non-exported fields — it's a one-shot signal payload, never persisted. This also sidesteps the scanner error that the prior @export var descriptor: CharacterVisualDescriptor on a Resource caused (RefCounted types cannot be @export-ed). - _on_start emits a CharacterProfile built from _descriptor + _selected_bookmark_id + _selected_location_id, then sends ConfirmBookmark via SimBridge.send_named_action before scene transition. Test updates: - test_character_creation_sprint28.gd signal receivers switched to untyped to accept CharacterProfile without hitting class_name parse-order at test-suite scan time. 88/88 pass. Verification: - gdlint clean - godot --headless --path client --quit — no SCRIPT ERROR (prior character_profile.gd scanner noise now gone after the RefCounted conversion) - test_character_creation_sprint28 88/88, test_protocol 62/62, test_implant_nav_stack 52/52 Workstream 7 (location picker as sub-component of the Bookmark detail view) follows. W8 fills the Skills tab. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>Review: sprint-36/client → main (PR #134)
Reviewers: Hoshe (code quality & tests) + Tyre (architecture).
Verdict: CHANGES REQUESTED
Four
error-severity issues block merge. Severalwarning/nititems recommended before close.Hoshe — Code Quality & Tests: REQUEST_CHANGES
MetaScreen migration is clean, triage skips carry ticket references, protocol v23 is symmetric. Two real bugs in character creation and one content-default nit in the protocol decoder.
client/ui/meta/screens/character_creation/character_creation.gd:36,196,~1613CARDINAL_DIRS = [south, west, north, east]vsCARDINAL_NAMES = [south, east, north, west]— both indexed by the same_screenshot_cardinal_idx. Screenshots at index 1 and 3 have swapped filename labels vs character facing. Either dropCARDINAL_NAMESentirely and derive the name fromCARDINAL_DIRS[idx], or document why the orderings differ.client/ui/meta/screens/character_creation/character_creation.gd:2034,2061KEY_ENTERbranch in_unhandled_inputcalls_on_start()without checking_footer_start.disabled— a player with no bookmark selected can press Enter to navigate into the game, sendingConfirmBookmarkwith empty strings. Addif _footer_start.disabled: returnguard at the top of_on_start().client/scripts/protocol/protocol.gd:412raw_bm.get("career", "tycoon")hardcodes a content default in the wire codec. A missing server field silently becomes"tycoon". Use""—_make_bookmark_cardalready skips the career label when empty.Tyre — Architecture: REQUEST_CHANGES
MetaScreen/MetaStack is a genuine abstraction win — lifecycle, ESC routing, and sim-pause coupling live in one place, with a clean division of labor vs HudGroups (MetaStack owns meta-UI stacking; HudGroups owns gameplay-vs-implant z-layering). Protocol v23 fits the snapshot model. But D-192's decision text disagrees with the code on this branch, and a few ESC-chain edges leak events across handlers.
decisions/architecture.md:752(D-192)PROTOCOL_VERSIONconstants and version-mismatch guard "are removed", but on this branch the constant is bumped to 23 and the guard indecode_snapshotis still active. Either reword the decision to "deprecated; removal tracked in #868" or land the removal in this PR. As written, the decision disagrees with the code on the same branch.client/ui/meta/meta_stack.gd:41-45handle_escape()returnsfalsefor a screen withclosable_by_escape = false(e.g.loading_screen) when itson_escape()default-returnsfalse.main.gdthen falls through to closing an implant app or opening settings — a screen declaring itself un-escapable should consume the event, not leak it. Returntruefrom thenot closable_by_escapebranch.client/ui/meta/screens/loading/loading_screen.gd:47"protocol %d" % Protocol.PROTOCOL_VERSION. Once D-192 lands this constant is gone; either drop the protocol-version line from the loading UI now (consistent with the decision) or soften the D-192 decision text. Same coupling as row 1.client/ui/meta/screens/debug_console/debug_console.gd:112KEY_ESCAPEdirectly in_unhandled_inputand callsclose(), bypassingMetaStack.handle_escape(). Second ESC handler living outside the documented pattern. Move the ESC branch out and rely on the MetaStack chain viaon_escape()override..claude/skills/pr-push/SKILL.mdclient/scripts/main.gd:63-70(Option A)GameState.bookmark_catalogsurvives the Option A scene transition via autoload — the dual-apply path (main_menu applies catalog → main scene re-applies on next tick) is non-obvious.client/ui/meta/screens/main_menu/main_menu.gd:66_ensure_loading_screen()instantiates a second LoadingScreen that is not the$MetaLayer/LoadingScreenmain.tscn uses. Two live instances share the same MetaStack autoload. Today works (main_menu.tscn and main.tscn never co-exist) but the "one LoadingScreen" invariant isn't enforced. Consider making LoadingScreen an autoload or document the lifecycle in the file header.client/ui/meta/meta_screen.gd:19-21captures_inputonly consulted inopen()to setmouse_filter, never unset inclose(). Asymmetric — if a screen setscaptures_input = falseinitially and later changes it, theopen()check wins forever. Apply both states symmetrically or drop the flag.client/ui/meta/meta_screen.gd:56-58on_escape()default returnsfalsewith no way to distinguish "consumed-and-held" from "did nothing, please close me" — MetaStack treats both the same. The tri-state protocol (consume-and-hold / consume-and-close / ignore) thathandle_escapesuggests isn't really expressible. Doc-comment clarifying thatclosable_by_escape = falseis the only way to say "consume-and-hold."client/scripts/main.gd:268(OPEN_MENU chain)_handle_menu_key()returning consumed/not, or a small ordered list.client/ui/meta/screens/bug_report/bug_report_dialog.gd:243on_escape()returns false (let MetaStack close) after emittingcapture_cancelled. Buton_close()does not emitcapture_cancelled— so if the dialog is closed via a non-ESC path, the signal is skipped. Move the emit intoon_close()unless completion has fired.Process note
PR body and commits don't mention a
make gameruntime smoke test. Given the ESC-chain and pre-game flow changes, a manual launch-through-character-creation run should happen before the next review round.Summary of blocking issues
4
error-severity fixes required:MetaStack.handle_escape()leaks events from un-escapable screensOther items are
warning/nit; recommend addressing before merge but may be bundled into a follow-up if the team pushes back with rationale.Addresses Hoshe's 3 code-quality items from the sprint-36 client review. - character_creation: drop CARDINAL_NAMES (was [south, east, north, west]) and use CARDINAL_DIRS ([south, west, north, east]) for both facing and screenshot filename label. The two arrays indexed by the same _screenshot_cardinal_idx produced swapped labels at indices 1 and 3 — screenshots at those positions had filenames that did not match the character's actual facing. - character_creation: Enter/KP_ENTER now honors _footer_start.disabled. Without a bookmark selected the Start button disables, but the keyboard path called _on_start() unconditionally — a player could confirm creation with empty bookmark/location strings. Guard at the top of _on_start. - protocol.gd: raw_bm.get("career", "tycoon") hardcoded a content default in the wire decoder — a missing server field silently became "tycoon". Empty string is the correct protocol default; _make_bookmark_card already skips the career label when empty.Review round 2 — all 14 items addressed
Pushed 3 commits (
effb83a0,1b8e2274,53fbce08) — all items from Hoshe + Tyre's review land in this push.Hoshe — Code Quality & Tests
effb83a0CARDINAL_NAMES, useCARDINAL_DIRSfor both facing and filename label. No more swapped labels at indices 1, 3.effb83a0_footer_start.disabledguard at top of_on_start()— Enter honors Start-button gating.effb83a0"career"default is now""instead of"tycoon".Tyre — Architecture
1b8e22741b8e2274MetaStack.handle_escape()consumes unconditionally forclosable_by_escape=falsescreens — ESC no longer leaks to the implant/settings chain behind a loading screen.loading_screenprotocol-version line is consistent. Will go in the #868 removal pass.1b8e2274debug_consoleno longer intercepts KEY_ESCAPE directly — falls through tomain.gd→MetaStacklike every other meta screen.1b8e2274main.gdnearconnect_to_simexplaining the Option A → GameState autoload catalog handoff.1b8e2274main_menu.gdheader documents the two-LoadingScreen lifecycle + conditions for promotion to autoload.1b8e2274captures_inputapplied symmetrically inMetaScreen.open()andclose().1b8e2274on_escape()docstring now spells out the tri-state (consume-and-hold / consume-and-close / ignore) and clarifies thatclosable_by_escape=falseis the screen-wide way to express consume-and-hold.1b8e2274_handle_menu_key()inmain.gd— ESC priority chain is now a named thing, not three inline early-returns.1b8e2274bug_report_dialogemitscapture_cancelledfromon_close()(covers all close paths). New_completedflag keepscapture_completedandcapture_cancelledmutually exclusive.Process note
tests/run-godothardened with a 300s hard timeout and a self-contained log-to-disk contract (53fbce08). Rationale: this round surfaced that the test runner was silently hanging for >10 min in Sprint 36 and blowing up caller context with 24k-line dumps. The script now:/tmp/sr-run-godot.log(always).TEST_TIMEOUTmarker and exit 124 — distinct from regular test failures.Scope-creep disclosure (Tyre #5)
Two cross-team files were edited on this client branch:
.claude/skills/pr-push/SKILL.md— hardened pre-push checks (orphan process check, widened Godot stderr grep). Committed earlier in sprint as804bba6a.tests/run-godot— the timeout/log hardening above. This round.Both are sprint-grounded (the exact failure modes bit this sprint), but neither is client-team scope. Flagging here for explicit sign-off rather than silently landing. Happy to split either into a separate PR against the appropriate branch if preferred.
Manual smoke test
Still pending. Will run
make game→ main menu → New Game → character creation tabs → bookmark picker end-to-end before requesting re-review and mark the test-plan checkboxes.Test health
Full
make test-clientpost-fixes: 1243 tests, 16 errors, 28 failures — identical to pre-review-fix state. All failures tracked in #869 (MetaScreen test-helper regression, sprint-36 caused), #870 (parse-error files), #871 (pre-existing umbrella). No new regressions from this review round.Review: sprint-36/client → main (PR #134) — Round 2
Reviewers: Hoshe (code quality & tests) + Tyre (architecture).
Verdict: CHANGES REQUESTED
One test-suite regression blocks merge. Tyre's 9 items are all addressed.
Hoshe — Code Quality & Tests: REQUEST_CHANGES
All 3 round-1 items correctly fixed. But the new
_on_startguard breaks 2 existing tests — the1b8e2274commit message explicitly flags the risk but leaves it unfixed.Round-1 status
CARDINAL_NAMESremoved,CARDINAL_DIRSused at every callsite, no residual references_on_start()guards on_footer_start.disabled; button and Enter both route through the guardcareerdefault hardcoded to"tycoon""";_make_bookmark_cardand the detail panel both guard on emptyNew round-2 comments
client/tests/test_character_creation_sprint28.gd:143-148, 460-468test_creation_confirmed_emits_on_startandtest_enter_emits_confirmedcall_on_start()/_input(KEY_ENTER)without seeding_selected_bookmark_id/_selected_location_id. With the new guard (_footer_start.disabledistruewhen both IDs are empty),creation_confirmednever fires andassert_signal(...).is_emitted(...)fails in any non-headless run. Fix: before the signal assert, set_scene._selected_bookmark_id = "test-bm"and_scene._selected_location_id = "test-loc"(or call_scene._update_start_btn_state()after seeding) so the button is enabled. The1b8e2274commit message flags this risk but doesn't resolve it.tests/run-godot:28LOG_FILE="/tmp/sr-run-godot.log"hardcoded. Concurrent runs across worktrees clobber each other's logs; the summary JSON may report counts from a different suite. The header comment says "concurrent runs are the caller's problem" — defensible, but/tmp/sr-run-godot.$$.logwould cost nothing and preserve the hint-line property by echoing the actual path used. Non-blocking.Tyre — Architecture: APPROVE
All 9 round-1 items landed correctly. The ESC contract is now explicit enough that a future screen author doesn't need tribal knowledge:
closable_by_escape = falsemeans "screen-wide consume-and-hold";on_escape() → truemeans "per-event consume-and-hold";on_escape() → falsemeans "let MetaStack pop me"._handle_menu_key()makes the priority chain a named thing a future fourth handler slots into rather than re-inlines.Round-1 status
PROTOCOL_VERSION=23, guard active, tautological asserts deleted). Merge conflict expected against main's independently-authored D-192 entry — flag for manual resolution, not a reviewer issuemeta_stack.gdESC leak on un-escapable screenshandle_escape()returnstrueunconditionally whenclosable_by_escape = falseloading_screen.gd:47PROTOCOL_VERSIONlabel couplingProtocol.PROTOCOL_VERSION; now covered by the deprecate-until-#868 framing. Acceptable — display-only, disappears when #868 landsdebug_console.gdsecond ESC handlerKEY_ESCAPEbranch removed; ESC falls through tomain.gd → _handle_menu_key() → MetaStack.handle_escape().claude/skills/pr-push/SKILL.mdscope creep804bba6astill carries the tooling edit. Not a blocker (round-1 warning); worth a note in PR bodymain.gdOption A bookmark_catalog autoload transition docGameState.bookmark_catalogsurvives the scene swap via autoloadmain_menu.gddouble-LoadingScreen invariantmeta_screen.gdcaptures_inputasymmetryopen()andclose()now apply the flag symmetricallymeta_screen.gdon_escape()tri-state undocumentedclosable_by_escapemain.gd:268ESC priority chain as early-returns_handle_menu_key()with a header comment explicitly naming the extension pathbug_report_dialog.gdcapture_cancelledemiton_close()gated by a new_completedflag; completion and cancel mutually exclusive across all close pathsNew round-2 comments
None blocking.
tests/run-godotrewrite is sound and composes cleanly with CI (JSON-on-stdout + hint-on-stderr + log-file). The cross-team scope-creep pattern (previously flagged forpr-pushSKILL) recurs here — acceptable given the self-contained improvement, but worth surfacing in the PR body alongside the SKILL.md edit.Merge conflict flag (not a reviewer issue)
Main's
decisions/architecture.mdnow contains a D-192 entry that was authored independently while the team was fixing their branch. Wording is similar but not identical. The branch's D-192 and main's D-192 will conflict on merge. Both texts are substantively compatible — resolve by hand taking whichever reads better.Summary of blocking issues
1
error-severity fix required:test_character_creation_sprint28.gd— 2 tests fail under the new_on_startguard because they don't seed a valid bookmark/location. Mechanical fix.All other round-1 items are FIXED (9) or PARTIALLY FIXED with acceptable rationale (1 — loading_screen label, covered by deprecate framing). One round-1 warning (pr-push SKILL scope creep) was not addressed and recurs in this round (
tests/run-godot) — not a merge blocker, but please call out both edits in the PR body for explicit sign-off.Review round 3 — all items addressed
Pushed
50ac9041with both Hoshe items resolved.Hoshe
test_character_creation_sprint28.before_eachnow seeds_selected_bookmark_id = "test-bookmark"and_selected_location_id = "test-location"and calls_update_start_btn_state(). Restores the 2 tests Hoshe flagged plus 3 more sibling tests that would have degraded the same way (every call site that hits_on_start()orKEY_ENTER). Tests that intentionally verify the disabled state can explicitly clear the IDs and call_update_start_btn_state()— the comment inbefore_eachspells this out. Isolated run: 88/88 ✅LOG_FILE="/tmp/sr-run-godot.$$.log"— PID suffix prevents concurrent runs across worktrees from clobbering each other's logs (and from emitting summary JSON that mixes counts from different suites). The actual path is still echoed back via the stdout JSONlogfield and the stderr hint line, so callers never need to predict it.Tyre
No changes needed — all 9 round-1 items stayed fixed after round 2, and the tests/run-godot rewrite holds up under the cross-team scope-creep caveat already surfaced in my round-2 comment.
Merge conflict on D-192
Fetched main; main does not currently have a D-192 entry (index
8540c7a0, tail still*53 decisions.).git merge-tree --write-tree origin/main HEADemits a clean tree, no conflict markers. The review flag was predictive — if an independent D-192 lands on main before this PR merges, it'll need hand-merging at that point, but there's nothing to resolve now.Test health
Full
make test-clientpost-fix: 1243 tests, 16 errors, 28 failures — identical baseline, no new regressions, no restored regressions beyond the 5 character-creation ones this round directly fixes. All remaining failures tracked in #869/#870/#871.Cross-team scope surface (for explicit sign-off)
Consolidated list of non-client-team files this PR touches, per Tyre #5:
.claude/skills/pr-push/SKILL.md804bba6a(sprint start)tests/run-godot53fbce08+50ac9041Both changes are self-contained and affect every team's
/pr-pushflow positively. Happy to split into a separate PR against a shared branch if preferred, but they ship cleanly as-is.Outstanding
make gamesmoke: failed — hangs on "connecting" after New Game. Filed as bug #872 (high priority, client team) with four ranked suspects and a repro recipe. Deferred to sprint 37 per review decision, not blocking merge.Review: sprint-36/client → main (PR #134) — Round 3
Reviewers: Hoshe (code quality & tests) + Tyre (architecture).
Verdict: APPROVED
Single round-3 commit (
50ac9041) cleanly addresses both R2 issues. All prior comments resolved across 3 rounds (14 → 1 → 0 blockers).Hoshe — Code Quality & Tests: APPROVE
Both R2 issues correctly resolved. One cosmetic nit, non-blocking.
Round-2 status
test_character_creation_sprint28.gd— tests call_on_start/Enter without seeding bookmark/locationbefore_eachseeds_selected_bookmark_id = "test-bookmark"and_selected_location_id = "test-location", then calls_update_start_btn_state()viahas_methodguard. No tests in the file assert the disabled state, so nothing needs to clear the seed; the comment inbefore_eachdocuments the pattern for future tests that would. Fix restores the 2 tests flagged + 3 siblings (_on_start/KEY_ENTER paths) that would have degraded the same way.tests/run-godot— hardcoded log path clobbered by concurrent runsLOG_FILE="/tmp/sr-run-godot.$$.log". All functional references (Godot redirect, stat parsing, JSONlogfield, stderr hint lines, timeout + failure branches) use$LOG_FILE.New comments
tests/run-godot:6"log":"/tmp/sr-run-godot.log"(pre-PID-suffix). Runtime uses$LOG_FILEcorrectly everywhere — this is stale documentation only. Update to"/tmp/sr-run-godot.<PID>.log"for accuracy. Can ride in a follow-up; does not block merge.Tyre — Architecture: APPROVE
PID-suffixed log path is applied consistently across declaration, both JSON branches, and both stderr hint branches — all use
$LOG_FILE.timeout --foreground --kill-after=10 $TIMEOUT_SECescalation and the stdout-JSON / stderr-hint contract are unchanged. No regressions.Observation (non-blocking)
One log per PID per run with no cleanup policy means
/tmp/sr-run-godot.*.logwill accumulate until/tmpis rotated by the OS. Not worth blocking on —/tmpcleanup is the platform's job and the files are small. If a future ticket adds a test-harness cleanup step, sweeping stalesr-run-godot.*.logolder than N days would fit naturally there.Pre-merge action (not a reviewer issue)
decisions/architecture.mdwill conflict on merge — both main and the branch have a D-192 entry with compatible-but-different wording. Resolve by hand; the branch's text is slightly more polished, take it.Summary
3-round review arc, 14 comments on round 1 → 1 blocker on round 2 → 0 blockers on round 3. MetaScreen pattern, ESC chain, protocol v23 bookmark catalog, and character-creation restructure all land cleanly. Approved for merge.
Pull request closed