Command-palette localization is the work of translating searchable commands, quick actions, launchers, command names, aliases, categories, parameters, shortcuts and result descriptions so users in every language find and execute the same operation. People searching for how to localize a command palette, translate quick actions, internationalize launchers or adapt searchable commands are solving both a language problem and a command-routing problem.
A professional multilingual launcher must survive synonyms, inflection, transliteration, non-Latin scripts, keyboard shortcuts, fuzzy matching, command aliases, dynamic permissions, recently used ranking, destructive actions, parameter prompts and platform conventions. A translation can look perfect but still be unsafe when the search layer returns the wrong command, two commands collapse to one label, or an alias in one language matches a destructive action unexpectedly.
This guide explains how to localize command palettes, quick actions and launchers without triggering the wrong command. It covers command IDs, display names, aliases, tokenization, fuzzy search, ranking, categories, parameters, shortcuts, context, permissions, destructive actions, accessibility, analytics, telemetry, locale switching and release QA.
The core rule is to treat command identity as immutable and search language as a locale-specific discovery layer. The palette can expose different words, synonyms and ordering by locale, but the selected result must still resolve to the same stable command, parameter schema and scope.
Good localization starts by documenting what every command does, where it is valid, what arguments it accepts, whether it is reversible and which permissions expose it. Once those semantics are stable, each locale can have natural names and search aliases without turning translation into application logic.
50-second router
- Give every command a stable machine ID independent of its displayed name.
- Localize command names, descriptions, categories and aliases as separate resources.
- Keep keyboard shortcuts as structured bindings, not translated prose.
- Use locale-aware tokenization and normalization for search.
- Do not let fuzzy matching outrank safety for destructive commands.
- Preserve command scope, permissions and parameter schemas.
- Test the same intent in source and target languages and compare command IDs.
- Log stable command IDs plus locale rather than rendered labels.
The central proposition
A command palette is safe to localize when discovery and execution are separate layers. Users may search with different words, scripts and synonyms, but a selected result should resolve through an invariant command ID to one well-defined action. That architecture lets language improve findability without giving translation the power to change what the software does.
1. Give commands stable IDs
The displayed command name is presentation, not the operation itself. Preserve command ID and execution handler while allowing localized name, description and aliases to adapt to the target locale. The main failure mode is that renaming a command can break routing, telemetry or automation when strings are used as keys. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Define invariant IDs such as file.open or view.toggleSidebar and map locale resources onto them. Example. Open File can become a natural target phrase while still resolving to file.open. Verification. Search and execute in several locales and compare the emitted command ID. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
2. Separate name from description
A short result name and a longer explanation serve different discovery needs. Preserve the command action and scope while allowing name length and descriptive phrasing to adapt to the target locale. The main failure mode is that translators can overload the short label with detail or omit a critical distinction. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Provide distinct resources for title and description with context on where each appears. Example. Close Workspace can have a description explaining that open files remain on disk. Verification. Inspect narrow and wide palette layouts and verify the distinction remains clear. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
3. Use locale-specific aliases
Users search with synonyms, abbreviations and alternate terminology. Preserve the command ID while allowing search aliases and keywords to adapt to the target locale. The main failure mode is that literal translation can make commands hard to find even though the label is correct. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Maintain locale-owned alias lists reviewed by people familiar with real target-language usage. Example. A command named Preferences may also match Settings in one locale without changing its canonical display label. Verification. Search each approved alias and confirm the same command ranks. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
4. Tokenize by language
Search systems built for spaces can fail in languages with different segmentation. Preserve indexed command identity while allowing tokenization, stemming and normalization to adapt to the target locale. The main failure mode is that a target command may become effectively undiscoverable. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Use locale-appropriate tokenization or language-aware search libraries where needed. Example. Japanese command names should not rely on English whitespace assumptions. Verification. Run a query set from native speakers and measure whether expected commands are retrieved. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
5. Normalize carefully
Case, accents, width variants and Unicode forms can affect matching. Preserve semantic text identity while allowing search normalization rules to adapt to the target locale. The main failure mode is that aggressive normalization can merge genuinely distinct strings or identifiers. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Normalize for discovery while retaining original display text and protected technical tokens. Example. Full-width and half-width forms may match in search while rendering exactly as translated. Verification. Test accent, case and Unicode-equivalent variants against expected command IDs. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
6. Treat transliteration as an optional bridge
Some users type Latin characters to find commands written in another script. Preserve the underlying command and native display name while allowing transliteration aliases to adapt to the target locale. The main failure mode is that automatic transliteration can produce collisions or culturally poor spellings. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Use transliteration as an additional discovery alias only when product evidence supports it. Example. A Cyrillic command may be discoverable through a Latin alias without replacing its native label. Verification. Search native and transliterated forms and inspect ambiguous results. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
7. Protect destructive commands in ranking
Delete, reset and revoke actions deserve conservative discovery. Preserve destructive class and confirmation policy while allowing localized verbs and search aliases to adapt to the target locale. The main failure mode is that fuzzy matching can surface a dangerous command for an imprecise query. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Apply safety-aware ranking and require strong matches or confirmation where appropriate. Example. Typing dele should not accidentally rank Reset Account above Delete Draft simply because of a translation artifact. Verification. Run near-miss query tests and inspect top results. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
8. Preserve command scope
Many actions apply to current file, selection, tab, workspace or account. Preserve target scope and execution context while allowing localized nouns and descriptions to adapt to the target locale. The main failure mode is that two commands can collapse into one phrase and hide which object will change. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Include scope in the command model and wording whenever ambiguity exists. Example. Close Tab and Close Window must remain distinguishable even if the target language uses similar verbs. Verification. Execute from different contexts and verify the intended object changes. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
9. Model parameters separately
Some commands require a path, user, branch, number or option after selection. Preserve parameter schema and validation while allowing prompt wording and examples to adapt to the target locale. The main failure mode is that translating the prompt cannot change the type or allowed values. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Define typed parameters and localize only names, help text and formatting. Example. Go to Line accepts an integer regardless of language while the prompt and number formatting can adapt. Verification. Enter valid and invalid values in each locale and compare validation. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
10. Keep enum values canonical
Parameters may offer modes or choices. Preserve stored enum value while allowing display label and ordering to adapt to the target locale. The main failure mode is that translated choices can be sent to APIs as if they were canonical values. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Map localized option labels to stable values. Example. A target label for Wrap Words still sends wrap_words to the command handler. Verification. Inspect execution payloads after choosing each localized option. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
11. Localize categories without changing hierarchy
Palettes often group commands by File, View, Git or Help. Preserve category ID and command membership while allowing category names and possibly display order to adapt to the target locale. The main failure mode is that renaming a category can accidentally become a new grouping in telemetry or search. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Store category membership using IDs and translate labels separately. Example. Commands in category view remain there even when the displayed category name changes. Verification. Compare category membership by ID across locales. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
12. Keep recent-use ranking language-independent
Frequently used commands may rise in results. Preserve usage history tied to command IDs while allowing display names and locale-specific query text to adapt to the target locale. The main failure mode is that switching language can erase personalization if history is stored by label. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Record usage against stable IDs and re-render the current locale. Example. A user’s frequent file.open command stays frequent after switching to Spanish. Verification. Switch locale on a profile with known history and compare ranked IDs. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
13. Keep search ranking explainable
Localized aliases and fuzzy scores can interact unpredictably. Preserve result command IDs and safety rules while allowing weights for locale resources to adapt to the target locale. The main failure mode is that a target-language synonym can outrank a more exact command unexpectedly. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Separate exact, prefix, alias and fuzzy match tiers and document their weights. Example. An exact match for Build should rank above a fuzzy alias of Rebuild Index unless product policy says otherwise. Verification. Run deterministic query fixtures per locale. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
14. Resolve duplicate labels deliberately
Different commands can legitimately translate to the same short phrase. Preserve distinct command IDs and scopes while allowing qualifiers in display text to adapt to the target locale. The main failure mode is that users may select the wrong result because two rows look identical. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Add concise contextual qualifiers or descriptions when target language collapses distinctions. Example. Open as Project and Open File may need different nouns in the translated result rows. Verification. Have reviewers choose from the ambiguous pair without icons, then refine until prediction is reliable. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
15. Keep shortcuts structured
A palette often shows key bindings alongside commands. Preserve binding and platform mapping while allowing localized key names and typography to adapt to the target locale. The main failure mode is that translating shortcut strings can invent unsupported keys. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Render shortcuts from structured key data according to platform conventions. Example. Command-P and Ctrl+P may both invoke the same palette command on different systems. Verification. Test real keyboard layouts and verify displayed hints match actual bindings. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
16. Localize access to the palette itself
The launcher trigger may have a menu label, tooltip or onboarding hint. Preserve the command that opens the palette while allowing instruction wording to adapt to the target locale. The main failure mode is that users may be told to press an unavailable shortcut or use source-language terminology. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Generate instructions from actual platform bindings and localize complete sentences. Example. Press Command-K to open Quick Actions should adapt to Windows if Ctrl-K is used there. Verification. Open the palette using every advertised method on each platform. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
17. Preserve permissions
Enterprise palettes often expose commands by role or capability. Preserve authorization rules while allowing role and command labels to adapt to the target locale. The main failure mode is that translation can be blamed for commands that should never have been visible. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Filter results using capability IDs before ranking localized text. Example. A viewer should not discover Publish even if they know its translated name. Verification. Search as multiple roles and compare result IDs. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
18. Handle disabled commands clearly
Some commands are visible but unavailable in current context. Preserve availability rule while allowing disabled-state explanation to adapt to the target locale. The main failure mode is that a localized label can imply the command is broken or generally forbidden. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Provide a localized reason from structured context state. Example. Rename Symbol may be disabled because no symbol is selected, not because the feature is unavailable. Verification. Change context until the command enables and verify the explanation changes correctly. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
19. Separate navigation from execution
Some palette entries open screens while others immediately modify state. Preserve command class and consequence while allowing verb choice and description to adapt to the target locale. The main failure mode is that a user may expect navigation and instead trigger an action. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Classify results as navigate, toggle, execute or prompt and write labels accordingly. Example. Open Settings navigates; Toggle Dark Mode changes state immediately. Verification. Ask target reviewers to predict the result type before activation. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
20. Localize toggle states correctly
Commands may represent current on/off state. Preserve setting ID and current value while allowing action wording and state description to adapt to the target locale. The main failure mode is that labels can describe the current state when they should describe the action, or vice versa. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Choose one product convention and expose state separately. Example. If the palette shows Enable Grid, it should not still say Enable after the grid is already on unless that is intentional. Verification. Toggle repeatedly and inspect both label and state announcement. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
21. Protect technical commands
Developer palettes may contain identifiers, paths and commands that must not be translated. Preserve literal token and execution syntax while allowing surrounding explanation to adapt to the target locale. The main failure mode is that translators can alter shell commands, flags or identifiers and break execution. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Mark protected spans and give translators context on what remains literal. Example. Run npm test can have a localized description while npm test stays unchanged. Verification. Execute copied examples from each locale build. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
22. Handle mixed-script queries
Users may search with project names, filenames and localized verbs together. Preserve protected dynamic text and command identity while allowing word order and bidi isolation to adapt to the target locale. The main failure mode is that mixed scripts can reorder punctuation or break matching. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Index dynamic tokens separately and use bidi-safe rendering. Example. Searching an Arabic verb plus an English filename should preserve the filename exactly. Verification. Test mixed RTL/LTR queries and inspect both matching and visual order. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
23. Make empty and no-match states useful
A palette can return nothing because of spelling, permissions or unavailable context. Preserve actual result set and reason while allowing localized guidance to adapt to the target locale. The main failure mode is that generic source wording can mislead users about why nothing appears. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Distinguish no lexical match from hidden-by-context where exposing that distinction is safe. Example. No commands found for this query is different from a disabled command with a known prerequisite. Verification. Create controlled no-result cases and review target-language guidance. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
24. Support IME composition
East Asian input methods produce intermediate composition text. Preserve composition lifecycle while allowing rendered candidate text to adapt to the target locale. The main failure mode is that search can execute or rerank before composition is committed. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Delay final action logic until composition events indicate committed input. Example. Typing a Japanese command name should not trigger an English shortcut mid-composition. Verification. Test real IME entry and rapid confirmation. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
25. Keep accessibility semantics explicit
Palettes behave like comboboxes, lists or dialogs with active descendants. Preserve roles, selected result and focus model while allowing localized names, instructions and result descriptions to adapt to the target locale. The main failure mode is that screen readers can announce stale labels or the wrong selected command. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Use established accessible patterns and update names from the same command resource as visible text. Example. The visually highlighted result and announced result should always share the same command ID. Verification. Navigate the palette with screen readers and compare highlight, announcement and execution. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
26. Log analytics by command ID
Teams study search terms and command usage. Preserve command ID and event type while allowing query text under appropriate privacy rules and locale to adapt to the target locale. The main failure mode is that analytics based on rendered labels fragment across languages. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Log invariant IDs and record locale separately; handle raw queries according to privacy policy. Example. file.open remains the event ID whether the visible label is English or Japanese. Verification. Inspect telemetry from identical actions in multiple locales. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
27. Test locale switching while open
Some applications can change language without restart. Preserve current query, selected command and command IDs while allowing rendered result labels and categories to adapt to the target locale. The main failure mode is that the highlighted row can point to a different command after re-sorting. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Recompute display and ranking while preserving selection by command ID where appropriate. Example. A selected file.open result should remain file.open after the interface language changes even if it moves position. Verification. Switch locale mid-palette and execute the selected result. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
28. Release with query-to-command fixtures
Visual review is not enough for searchable command systems. Preserve expected command IDs for representative queries while allowing localized query phrases and aliases to adapt to the target locale. The main failure mode is that a subtle search change can route target users to different operations. Localization should therefore begin from interaction semantics, not from isolated source wording.
Professional method. Maintain per-locale fixtures from common and safety-sensitive intents. Example. Each fixture states the query, expected top IDs and commands that must not appear above them. Verification. Run fixtures on every release and review deviations with localization and product teams. Repeat the check after switching locale, resizing the interface, changing input method and restoring saved state so that the behavior remains stable under realistic multilingual use.
Worked failure-and-repair scenarios
1. Two commands translate to the same label
Add scope or context to the display labels while retaining distinct IDs and execution handlers. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
2. A fuzzy query surfaces Delete Account
Strengthen safety-aware ranking and require a more deliberate match or confirmation path for destructive operations. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
3. A user changes language and recent commands disappear
Store usage history by command ID and re-render localized names from that stable history. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
4. An alias in one locale matches the wrong command
Review locale alias collisions and separate exact, prefix and fuzzy tiers. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
5. IME composition triggers a shortcut early
Respect composition events and defer command execution until input is committed. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
6. A parameter label is translated and sent to the API
Map display labels back to canonical enum values before execution. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
7. A viewer can search for an admin-only action
Apply permissions before indexing or result display rather than hiding only the English label. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
8. A shortcut hint is correct on macOS but wrong on Windows
Render bindings from platform-specific structured data, not copied strings. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
9. A right-to-left query visually scrambles a filename
Use bidi isolation for protected dynamic tokens and test mixed-script rendering. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
10. A command stays selected after re-ranking but points to another row
Track selection by command ID rather than list index. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
11. No results text implies the feature does not exist
Distinguish lexical no-match from contextual unavailability where doing so is safe. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
12. Telemetry shows separate events for every language
Replace label-based event names with stable command IDs and locale metadata. Record the invariant first, reproduce the failure in at least two locales, repair the earliest layer that caused it, and retest the same interaction using keyboard, pointer or touch as appropriate. This prevents a cosmetic translation patch from hiding a deeper state or layout defect.
Release checklist
- Every command has a stable machine ID.
- Names, descriptions, aliases and categories are separate localizable resources.
- Locale-aware tokenization and normalization are configured.
- Safety-sensitive commands have deliberate ranking rules.
- Command scope is explicit.
- Parameter schemas remain typed and locale-independent.
- Enum values map from localized labels to canonical values.
- Recent history is stored by command ID.
- Duplicate target labels are disambiguated.
- Shortcuts render from structured bindings.
- Permissions filter results before execution.
- Disabled reasons are localized from structured state.
- IME composition is handled safely.
- Accessibility selection matches the visible result.
- Analytics use invariant IDs.
- Per-locale query fixtures pass before release.
Frequently asked questions
Should command IDs ever be translated?
No. Command IDs are machine identity. Translate the display name, description, category and aliases around them.
How many aliases should a locale have?
Enough to cover common real-world terminology without creating noisy collisions. Native-speaker query data is more useful than mechanically translating a long source alias list.
Can fuzzy matching be the same in every language?
The overall method can be shared, but tokenization, normalization and weighting often need locale-aware behavior.
How should destructive commands appear in a palette?
Use clear target-language wording, conservative matching and the same confirmation policy as elsewhere in the product.
What about keyboard shortcuts?
Treat them as structured bindings tied to platform and keyboard behavior. Localize key names and help text separately.
How do I test a multilingual command palette?
Use locale-specific query fixtures, compare returned command IDs, test parameters, permissions, shortcuts, IME input and screen-reader navigation.
Should recent commands reset after language change?
Usually no. Usage history belongs to command identity, not to the current rendered label.
Which SEO phrases are relevant?
Useful terms include command palette localization, quick action translation, launcher internationalization, multilingual fuzzy search, localized command aliases and command palette accessibility.
Final principle
Localize how users discover a command, not what the command is. When names, aliases, categories and search behavior sit on top of stable command IDs, scope, permissions and parameters, every language can feel natural without changing execution. That separation is the foundation of a safe multilingual command palette.