Final verification for c2caabd26d8c2a50daa6f061e78449aa233d8d5e:

• Added a regression test for deleting a modifier-only Primary Dictation shortcut while its hold start is pending.
• Local Xcode 26.6: swiftlint --strict --config .swiftlint.yml passed with 0 violations.
• Fork CI on the same head passed: https://github.com/Yaomeng1749/FluidVoice/actions/runs/28551228981
• CI covered SwiftLint, dependency resolution, build, and tests.
• Upstream Actions are still waiting for maintainer approval for the public-fork workflow.

Thanks for the PR. Actually, very useful PR, and thanks for finding the bug as well, but I think there is one comment that I have for you. Let me know what you think. If you think this is valid, then you can probably fix it for me and then you can merge it .

"active primary press matching only uses keyCode/button. If two primary shortcuts share the same key/button with different modifiers, deleting the inactive one can clear the active held one, so release may not stop recording. Fix: store/compare the active HotkeyShortcut including modifiers."

Thanks, I think this is valid.

The current cleanup path identifies the active primary press by key/button only, which is fine for pairing the release event, but too broad for deciding whether a removed shortcut is the active one. With two primary shortcuts sharing the same key/button but different modifiers, removing the inactive one could clear the active press.

I’ll update the active primary press state to keep the matched HotkeyShortcut and use that full shortcut, including modifiers, when comparing against removed shortcuts. I’ll keep the release path tolerant of modifier changes on key-up, but make the removal cleanup use the full shortcut identity. I’ll also add a regression test for the shared key/button + different modifiers case.

Implemented this in 55a469d2946f3c0f8d2837476cb4aa4731afb2b0.

The active primary press now snapshots the matched shortcut identity, including modifiers, for removal matching. Key-up / mouse-up still only match the released key/button so releasing modifiers before the main key remains supported.

I also added a regression test for the shared-key case: Command+K active, Option+K removed. The active press is preserved and the later K release clears it.

Local checks:

• swiftformat --lint --config .swiftformat Sources/Fluid/Services/GlobalHotkeyManager.swift Tests/FluidDictationIntegrationTests/HotkeyShortcutTests.swift
• git diff --check
• xcrun swiftc -parse Sources/Fluid/Services/GlobalHotkeyManager.swift Tests/FluidDictationIntegrationTests/HotkeyShortcutTests.swift
• swiftlint --strict --config .swiftlint.yml
• targeted xcodebuild test for FluidDictationIntegrationTests/GlobalHotkeyManagerShortcutTests: 2 passed, 0 failed on macOS/My Mac.