Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Maintain DomoDomo’s dark visual system, responsive layouts, accessible interactions, brand consistency, and component ergonomics. Use when changing styles, shared components, pages, forms, overlays, controls, or visual tool workspaces.
DomoDomo Design and Accessibility
Maintain The visual and interaction system used by all DomoDomo tools, pages, and overlays across desktop and mobile browsers.
Operating contract
Scope: The visual and interaction system used by all DomoDomo tools, pages, and overlays across desktop and mobile browsers.
Primary files: src/index.css, src/App.css, src/utils/BrandKit.ts, src/components/, src/pages/, and UI-heavy files in src/tools/.
Before changing code: Read BrandKit, shared CSS, a comparable responsive component, and the current accessibility behavior before introducing visual patterns.
01. brand color usage
For brand color usage, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For brand color usage, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For brand color usage, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For brand color usage, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For brand color usage, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For brand color usage, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For brand color usage, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For brand color usage, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For brand color usage, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For brand color usage, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For brand color usage, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For brand color usage, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For brand color usage, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For brand color usage, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For brand color usage, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For brand color usage, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For brand color usage, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For brand color usage, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For brand color usage, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For brand color usage, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For brand color usage, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
02. typography hierarchy
For typography hierarchy, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For typography hierarchy, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For typography hierarchy, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For typography hierarchy, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For typography hierarchy, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For typography hierarchy, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For typography hierarchy, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For typography hierarchy, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For typography hierarchy, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For typography hierarchy, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For typography hierarchy, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For typography hierarchy, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For typography hierarchy, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For typography hierarchy, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For typography hierarchy, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For typography hierarchy, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
03. glass-card composition
For glass-card composition, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For glass-card composition, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For glass-card composition, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For glass-card composition, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For glass-card composition, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For glass-card composition, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For glass-card composition, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For glass-card composition, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For glass-card composition, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For glass-card composition, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For glass-card composition, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For glass-card composition, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For glass-card composition, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For glass-card composition, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For glass-card composition, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For glass-card composition, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
04. spacing rhythm
For spacing rhythm, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For spacing rhythm, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For spacing rhythm, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For spacing rhythm, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For spacing rhythm, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For spacing rhythm, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For spacing rhythm, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For spacing rhythm, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For spacing rhythm, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For spacing rhythm, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For spacing rhythm, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For spacing rhythm, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For spacing rhythm, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For spacing rhythm, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For spacing rhythm, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For spacing rhythm, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For spacing rhythm, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
05. responsive breakpoints
For responsive breakpoints, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For responsive breakpoints, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For responsive breakpoints, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For responsive breakpoints, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For responsive breakpoints, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For responsive breakpoints, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For responsive breakpoints, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For responsive breakpoints, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For responsive breakpoints, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For responsive breakpoints, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For responsive breakpoints, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For responsive breakpoints, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For responsive breakpoints, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For responsive breakpoints, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For responsive breakpoints, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For responsive breakpoints, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
06. mobile layout behavior
For mobile layout behavior, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For mobile layout behavior, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For mobile layout behavior, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For mobile layout behavior, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For mobile layout behavior, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For mobile layout behavior, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For mobile layout behavior, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For mobile layout behavior, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For mobile layout behavior, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For mobile layout behavior, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For mobile layout behavior, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For mobile layout behavior, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For mobile layout behavior, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For mobile layout behavior, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For mobile layout behavior, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For mobile layout behavior, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
07. keyboard navigation
For keyboard navigation, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For keyboard navigation, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For keyboard navigation, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For keyboard navigation, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For keyboard navigation, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For keyboard navigation, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For keyboard navigation, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For keyboard navigation, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For keyboard navigation, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For keyboard navigation, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For keyboard navigation, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For keyboard navigation, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For keyboard navigation, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For keyboard navigation, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For keyboard navigation, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For keyboard navigation, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
08. focus visibility
For focus visibility, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For focus visibility, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For focus visibility, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For focus visibility, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For focus visibility, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For focus visibility, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For focus visibility, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For focus visibility, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For focus visibility, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For focus visibility, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For focus visibility, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For focus visibility, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For focus visibility, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For focus visibility, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For focus visibility, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For focus visibility, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For focus visibility, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
09. semantic element choice
For semantic element choice, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For semantic element choice, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For semantic element choice, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For semantic element choice, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For semantic element choice, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For semantic element choice, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For semantic element choice, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For semantic element choice, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For semantic element choice, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For semantic element choice, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For semantic element choice, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For semantic element choice, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For semantic element choice, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For semantic element choice, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For semantic element choice, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For semantic element choice, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
10. form label clarity
For form label clarity, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For form label clarity, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For form label clarity, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For form label clarity, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For form label clarity, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For form label clarity, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For form label clarity, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For form label clarity, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For form label clarity, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For form label clarity, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For form label clarity, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For form label clarity, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For form label clarity, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For form label clarity, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For form label clarity, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For form label clarity, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For , inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
11. error message association
For error message association, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For error message association, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For error message association, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For error message association, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For error message association, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For error message association, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For error message association, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For error message association, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For error message association, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For error message association, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For error message association, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For error message association, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For error message association, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For error message association, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For error message association, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For error message association, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
12. dialog and overlay behavior
For dialog and overlay behavior, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For dialog and overlay behavior, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For dialog and overlay behavior, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For dialog and overlay behavior, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For dialog and overlay behavior, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For dialog and overlay behavior, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For dialog and overlay behavior, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For dialog and overlay behavior, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For dialog and overlay behavior, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For dialog and overlay behavior, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For dialog and overlay behavior, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For dialog and overlay behavior, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For dialog and overlay behavior, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For dialog and overlay behavior, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For dialog and overlay behavior, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For dialog and overlay behavior, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
13. tooltip behavior
For tooltip behavior, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For tooltip behavior, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For tooltip behavior, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For tooltip behavior, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For tooltip behavior, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For tooltip behavior, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For tooltip behavior, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For tooltip behavior, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For tooltip behavior, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For tooltip behavior, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For tooltip behavior, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For tooltip behavior, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For tooltip behavior, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For tooltip behavior, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For tooltip behavior, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For tooltip behavior, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For tooltip behavior, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
14. icon-only controls
For icon-only controls, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For icon-only controls, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For icon-only controls, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For icon-only controls, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For icon-only controls, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For icon-only controls, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For icon-only controls, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For icon-only controls, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For icon-only controls, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For icon-only controls, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For icon-only controls, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For icon-only controls, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For icon-only controls, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For icon-only controls, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For icon-only controls, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For icon-only controls, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For , inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
15. color contrast
For color contrast, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For color contrast, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For color contrast, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For color contrast, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For color contrast, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For color contrast, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For color contrast, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For color contrast, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For color contrast, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For color contrast, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For color contrast, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For color contrast, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For color contrast, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For color contrast, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For color contrast, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For color contrast, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For color contrast, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
16. motion preferences
For motion preferences, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For motion preferences, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For motion preferences, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For motion preferences, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For motion preferences, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For motion preferences, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For motion preferences, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For motion preferences, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For motion preferences, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For motion preferences, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For motion preferences, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For motion preferences, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For motion preferences, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For motion preferences, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For motion preferences, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For motion preferences, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
For , inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
17. canvas control accessibility
For canvas control accessibility, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For canvas control accessibility, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For canvas control accessibility, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For canvas control accessibility, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For canvas control accessibility, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For canvas control accessibility, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For canvas control accessibility, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For canvas control accessibility, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For canvas control accessibility, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For canvas control accessibility, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For canvas control accessibility, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For canvas control accessibility, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For canvas control accessibility, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For canvas control accessibility, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For canvas control accessibility, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For canvas control accessibility, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
18. screen-reader announcements
For screen-reader announcements, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For screen-reader announcements, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For screen-reader announcements, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For screen-reader announcements, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For screen-reader announcements, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For screen-reader announcements, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For screen-reader announcements, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For screen-reader announcements, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For screen-reader announcements, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For screen-reader announcements, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For screen-reader announcements, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For screen-reader announcements, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For screen-reader announcements, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For screen-reader announcements, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For screen-reader announcements, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For screen-reader announcements, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
19. visual regression checks
For visual regression checks, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For visual regression checks, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For visual regression checks, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For visual regression checks, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For visual regression checks, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For visual regression checks, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For visual regression checks, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For visual regression checks, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For visual regression checks, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For visual regression checks, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For visual regression checks, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For visual regression checks, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For visual regression checks, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For visual regression checks, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For visual regression checks, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For visual regression checks, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
20. design change completion
For design change completion, establish the exact user-facing outcome and explicitly state which behavior is intentionally out of scope.
For design change completion, read the nearest existing implementation before creating files, types, helpers, or a second pattern for the same job.
For design change completion, map every caller, consumer, and persisted or exported output affected by this area before changing a contract.
For design change completion, prefer the smallest local implementation that fits the established DomoDomo architecture and preserves offline behavior.
For design change completion, use explicit TypeScript or Python types at boundaries; reject malformed values early and explain recovery to the user.
For design change completion, keep user files, prompts, tokens, metadata, and derived artifacts on the device unless a documented local service is required.
For design change completion, avoid adding a remote dependency, analytics call, credential requirement, or hidden network transfer to solve a local workflow.
For design change completion, design empty, invalid, loading, cancellation, permission-denied, and failure states before treating the happy path as complete.
For design change completion, release resources deterministically: revoke object URLs and terminate workers, listeners, streams, timers, and subscriptions.
For design change completion, bound memory, CPU, and rendering work for large inputs; provide progress and a clear limit instead of freezing the interface.
For design change completion, reuse DomoDomo visual patterns, semantic controls, labels, focus behavior, responsive rules, and accessible feedback.
For design change completion, keep identifiers stable and align names, categories, descriptions, paths, exports, and documentation across all affected surfaces.
For design change completion, record non-obvious browser, model, file-format, hardware, or platform limitations where agents and users will encounter them.
For design change completion, treat logs, exception messages, generated artifacts, and command output as potentially sensitive; minimize what is retained or shown.
For design change completion, verify that retries and repeated actions are safe and do not duplicate records, leak handles, or leave stale UI state.
For design change completion, test the narrowest realistic scenario first, then test a large or slow input and a user interruption before declaring success.
Completion standard
Confirm every changed surface follows this playbook and its local-first constraints.
Confirm applicable documentation, About metrics, technical records, and blog content accurately describe the final behavior.
Confirm the handoff records the commands run, validation outcome, known limits, and intentionally deferred work.
For typography hierarchy, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For typography hierarchy, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For typography hierarchy, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For typography hierarchy, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For typography hierarchy, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For glass-card composition, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For glass-card composition, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For glass-card composition, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For glass-card composition, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For glass-card composition, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For spacing rhythm, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For spacing rhythm, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For spacing rhythm, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For spacing rhythm, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For responsive breakpoints, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For responsive breakpoints, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For responsive breakpoints, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For responsive breakpoints, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For responsive breakpoints, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For mobile layout behavior, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For mobile layout behavior, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For mobile layout behavior, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For mobile layout behavior, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For mobile layout behavior, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For keyboard navigation, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For keyboard navigation, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For keyboard navigation, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For keyboard navigation, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For keyboard navigation, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For focus visibility, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For focus visibility, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For focus visibility, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For focus visibility, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For semantic element choice, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For semantic element choice, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For semantic element choice, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For semantic element choice, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For semantic element choice, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
form label clarity
For form label clarity, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For form label clarity, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For form label clarity, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For form label clarity, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For error message association, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For error message association, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For error message association, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For error message association, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For error message association, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For dialog and overlay behavior, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For dialog and overlay behavior, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For dialog and overlay behavior, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For dialog and overlay behavior, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For dialog and overlay behavior, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For tooltip behavior, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For tooltip behavior, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For tooltip behavior, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For tooltip behavior, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
icon-only controls
For icon-only controls, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For icon-only controls, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For icon-only controls, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For icon-only controls, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For color contrast, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For color contrast, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For color contrast, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For color contrast, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
motion preferences
For motion preferences, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For motion preferences, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For motion preferences, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For motion preferences, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For canvas control accessibility, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For canvas control accessibility, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For canvas control accessibility, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For canvas control accessibility, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For canvas control accessibility, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For screen-reader announcements, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For screen-reader announcements, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For screen-reader announcements, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For screen-reader announcements, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For screen-reader announcements, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For visual regression checks, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For visual regression checks, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For visual regression checks, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For visual regression checks, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For visual regression checks, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.
For design change completion, inspect the final diff for accidental generated files, credentials, databases, model data, unrelated refactors, and false claims.
For design change completion, run the applicable project command and report its result; do not infer correctness from a visual inspection alone.
For design change completion, update product-facing documentation, technical records, metrics, and blog content whenever this area changes a public capability.
For design change completion, preserve unrelated working-tree changes and do not rewrite another contributor’s implementation without explicit scope.
For design change completion, finish with a concise handoff that identifies changed files, verification evidence, limitations, and remaining follow-up work.