feat(loading-skeleton): responsive sizing across mobile/tablet/desktop - #349

…wallet_badge

- Replace hardcoded empty message with EmptyStateCard component
- Display briefcase icon for job-related context
- Show descriptive title and explanation text
- Include role badges (Client, Freelancer, Arbiter) showing available participation options
- Add comprehensive test coverage with 16 test cases
- Ensure proper accessibility with region landmarks and aria-labels
- Validates placeholder display under empty data states

…ldii-locks#310)

…ldii-locks#311)

…i-locks#313)

…i-locks#318)

…i-locks#311)

…#313)

…#318)

Implements the app dark/light theme toggle as a keyboard-operable,
ARIA-compliant switch (role=switch, aria-checked, aria-label) that
persists the chosen theme to localStorage and applies it to the document
root. Adds React Testing Library tests verifying node rendering,
accessible state, theme application, keyboard operation and persistence.

Closes Goldii-locks#319

Implements the navbar alert bell badge as a keyboard-operable
(role=button) disclosure with ARIA compliance (Goldii-locks#320): accessible name,
aria-haspopup/aria-expanded/aria-controls, aria-live announcement
regions, aria-hidden on decorative glyphs, focus-visible rings and
design-token contrast.

Adds field error indicators and alerts (Goldii-locks#324): validation field configs
render role=alert error text that toggles as validation triggers, wired
via aria-describedby and counted toward the unread badge.

Adds React Testing Library tests covering both requirements.

Closes Goldii-locks#320
Closes Goldii-locks#324

Stack the stat grid and content rows to a single column on mobile,
two on tablet, three on desktop, and use responsive padding/spacing
so LoadingSkeleton scales cleanly at every breakpoint.

Closes Goldii-locks#275

Replace the bare "No jobs found" line with a reusable EmptyState
component (icon, heading, supporting copy) shown once loading
finishes with zero jobs for the connected wallet.

Closes Goldii-locks#276

…ad of setState-in-effect

CI's lint step has been red since Goldii-locks#283/Goldii-locks#284/Goldii-locks#287: ESLint's
react-hooks/set-state-in-effect rule flags the effect that called
setParseMessage synchronously right after checking activeTransactionXdr.
The value is purely derived from activeTransactionXdr, so compute it
with useMemo during render instead of scheduling a second render via
an effect.

The two new test files relied on ambient describe/it/expect and on
jest.fn()/jest.spyOn(). This project runs vitest without globals typing
for these paths, so tsc failed on every assertion and the spies had no
implementation. Imported the vitest helpers explicitly and switched the
four jest.* calls to their vi.* equivalents.

…allet-disconnect-handler-Goldii-locks#242-FIX

…gas-warning

 feat: Add gas estimation warning banners to network_sync_checker

…ution)

The PR imported EmptyStateCard but never added the component, so the
dashboard would not build. Added it, satisfying this PR's contract
(dashboard-empty-state test id, region role with an aria-label, an SVG
briefcase icon and role badges) while also carrying the shared
empty-state/-title/-description ids the existing dashboard empty-state
test queries.

The description now carries both wordings so the existing assertion
(/create one to get started/i) and this PR's fuller copy both hold.

…allet-disconnect-handler-Goldii-locks#242-FIX

…y-state-ui

feat: implement descriptive empty state UI for dashboard jobs list

…torybook-318

Build Storybook interface mocks for dark_mode_switcher

The branch was pushed with a botched conflict resolution: the `>>>>>>>`
markers were stripped but the branch-name lines and both sides of each
conflict were left in place. That left four stray
`Write-React-Testing-Library-assertions-...-Goldii-locks#242-FIX` lines, a duplicate
`@stellar/stellar-sdk` import block, a second `parseMessage` useMemo in
SignatureTimeoutAlert, and a duplicate `server` key in vitest.config.mts.

Resolved by taking main for the five files the branch only touched
incidentally -- main already carries the equivalent, deduplicated work
(the same parseMessage useMemo derivation and the same freighter-api
inline config the branch was re-adding).

The branch's actual deliverable for Goldii-locks#242 is kept intact:
WALLET_DISCONNECT_HANDLER_TESTS.md and the 851-line
wallet_disconnect_handler.component.test.ts.

lint 0 errors / tsc 0 errors / 1776 tests passing / build OK

…Testing-Library-assertions-for-wallet-disconnect-handler-#242-FIX

Write React Testing Library assertions for wallet disconnect handler …

…ns target

The two design-token cases rendered WalletSelectorModal with its default
props, where `errorMessage` defaults to null and `activeAddress` defaults
to null. Both surfaces under test are gated on those props, so
`wallet-selector-error-message` and `wallet-selector-connected-badge`
were never in the tree and the queries threw.

Passed an errorMessage to the first case and an activeAddress to the
second. The component's design tokens were already correct -- only the
test setup needed the props.

lint 0 errors / tsc 0 errors / 1778 tests passing / build OK

…e-modal-design-tokens

Feat: Dispute Raise ModaL

…n tokens

The suite hardcoded the raw Tailwind palette LoadingSkeleton used when
the branch was written. main has since migrated the component to
semantic design tokens, so the class assertions and the querySelector
calls that locate the cards no longer matched anything.

Remapped throughout:
  bg-gray-900                  -> bg-surface-card
  bg-gray-800 (bars, stat card)-> bg-surface-field
  bg-gray-700 (stat placehold.)-> bg-border-subtle
  border-gray-800              -> border-border-strong

Worth noting three cases were passing vacuously: the stat- and
milestone-card selectors returned empty NodeLists, so their forEach
assertion bodies never ran. With the selectors fixed they now iterate
and pass for real -- 54 of 54 in this file, up from 49.

lint 0 errors / tsc 0 errors / 1832 tests passing / build OK

…ng-spinner-skeleton

test: Add React Testing Library tests for LoadingSkeleton and ButtonSpinner (Goldii-locks#282)

… design tokens

The branch was cut before main migrated LoadingSkeleton to semantic
design tokens and added the skeleton-* test ids, so its side of the
conflict still carried bg-gray-*/no-testid markup. Resolved by applying
the branch's responsive layout on top of main's tokenised markup rather
than taking either side whole.

Two adjustments to keep the sibling suites passing:

- The branch's new test ids (loading-skeleton-card/-stats/-rows) name
  elements that already carry ids on main (skeleton-container,
  skeleton-stats-grid, skeleton-milestones). An element can only hold
  one data-testid, so the test now queries main's names -- every
  assertion is unchanged.

- Placeholder bars are expressed desktop-first (w-32 max-sm:w-24) rather
  than mobile-first (w-24 sm:w-32). Same computed result at both
  breakpoints, but it keeps the fixed widths on the element so the
  structural selectors in loading-spinner-skeleton.test.tsx still match.

The two card-padding assertions there did have to move to the responsive
class (p-6 -> sm:p-6, space-y-6 -> sm:space-y-6), since the branch
deliberately makes those breakpoint-dependent.

lint 0 errors / tsc 0 errors / 1841 tests passing / build OK