[malpern]

Replace, not wrap (review §4.4g).

What

• AdmissionEngine.conflicts(in:leaderKey:) computes every conflict from claims — what each owner puts on which slot — attributed by rule id. Five kinds of collision in one pass: key slot (with the leader key's base slot, fix: detect leader key vs collection activator conflicts #463), activator shadowing a mapping (fix: make momentary activator dedup detect-and-explain instead of silent #466), two activators disagreeing on a target layer, chord groups within one owner (fix: detect duplicate chord group names #464), and alias-name collisions (fix: detect alias name collisions from key sanitization #462).
• Reason.conflict gains details (hold action, "activates X", alias origin); the presenter and the save gate's MappingConflictInfo render the same text.
• RuleCollectionDeduplicator.detectConflicts is deleted. The save gate (ConfigurationService.generateConfiguration) and the CLI (RulesFacade) call AdmissionEngine.mappingConflicts(in:leaderKey:), the nonisolated bridge to the error type. The deduplicator keeps only dedupe.
• Fixes the name-based attribution: two rules titled alike, one colliding, now blame only the one that collides.

Tests

• The old detector's entire corpus (RuleCollectionDeduplicatorTests, MultiCollectionConflictTests, AliasDeduplicationTests, DeduplicatorEdgeCaseTests, ChordGroupsIntegrationTests, CLI and save-pipeline tests) now runs against the engine, unchanged in its assertions — the proof the replacement preserves behaviour.
• New: testSameTitledRulesAreToldApart, testConflictDetailsNameTheHoldAndTheActivator.
• KEYPATH_TEST_ENFORCE_CLEAN_SUMMARY=1 Scripts/run-tests-safe.sh: 5,786 passed, 0 app warnings/errors, 0 build warnings. check-accessibility.py clean.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-20T12:18:24.641755Z	ba5578f	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[claude]

Review of PR #1322 (the diff was truncated at 60 KB, so I only saw part of the test changes).

1. Chord-group conflicts now have no fixes (AdmissionEngine.swift, section 4 of conflicts(owners:leaderKey:))

• Both chord checks (duplicate group names and cross-group shared keys) use .named(...) parties. conflict(...) drops .named parties when it builds fixes.
• So these .needsChoice verdicts now have fixes == []. The old code offered .disable(.named(...)) for each party, which wasn't actionable either. Now the UI gets a blocking verdict with no option to resolve it.
• RuleCollectionsManager+ConflictDetection matches on parties.contains(subject). The owning collection is no longer among the parties, so a toggle-time check won't attribute these conflicts to the chord collection.
• Suggested fix: include .disable(owner.subject) in the fixes for these verdicts, and consider adding owner.subject to parties. A test asserting that fixes are non-empty for chord conflicts would cover it.

2. Doc comment typo (KanataConfiguration+BlockBuilders.swift)

• It reads AdmissionEngine.conflicts(in:leaderKey:)(), with stray parentheses. Also, the primary gate is really mappingConflicts, which the save path calls.

3. Minor

• mappingConflicts falls back to "a rule" when an id isn't in collections. Two different unknown parties then collapse to the same label.
• The @MainActor enum now exposes nonisolated static entry points that call KanataConfiguration.effectiveMappings and generatedAliasNames. Please confirm those are nonisolated. Otherwise this won't compile under strict concurrency, or it will hop actors.