{
  "intent": "Bring kcd2_dice_solver/src/App.tsx (306 lines) under the repo-wide 250-line cap with no change to what the app renders or how the inventory behaves, completing the four-file kcd2 series.",
  "scope": [
    "kcd2_dice_solver/src/App.tsx — reduced to the view: props, the search query, the solve request, and the JSX",
    "kcd2_dice_solver/src/hooks/useInventoryState.ts — new: InventoryState, useInventoryState",
    "Non-goal: any change to rendered markup, class names, ARIA roles or copy, which the four App test files query on",
    "Non-goal: a dedicated test file for the hook, which would change the 288-test count this split is verified against"
  ],
  "changes": [
    "Extracted the inventory state into a useInventoryState hook under src/hooks/, following the existing useSolver pattern: the link-versus-storage resolution at mount, the deferred-persistence boolean and its storage effect, the fragment-clearing effect, and the setDieCount/toggleBadge mutators, all with their explanatory comments.",
    "Promoted the two inline handlers that only set state — Keep, and the Discard/Clear/import replacement — to named `keep` and `replace` callbacks on the hook, so the view no longer touches setPersist or setSaved directly. importInventory collapsed into `replace`, which Clear, Discard and InventoryIo's onImport now all use.",
    "The hook exposes `stored` as a plain SavedInventory rather than the loadInventory result, since Discard is the only consumer and it wants the inventory, not the parse report.",
    "Deliberately avoided a re-export: an initial draft had the hook re-export EMPTY, which is the barrel pattern this series is not using. App.tsx imports EMPTY from lib/inventoryIo.ts, its real home."
  ],
  "verification": [
    {
      "command": "npx vitest run",
      "result": "pass",
      "evidence": "Test Files 24 passed (24); Tests 288 passed (288). The four App test files — AppStorage, AppShell, AppLinks, AppRecovery — exercise the extracted state through the component and all still pass unmodified."
    },
    {
      "command": "npx vitest run --coverage",
      "result": "pass",
      "evidence": "Statements 100% (864/864), Branches 100% (375/375), Functions 100% (182/182), Lines 100% (811/811). Totals rose by exactly one statement and one function against the 863/375/181/810 baseline, which is the Keep handler becoming a named callback; branches are unchanged, so no conditional left test reach. This was the metric the brief warned would drop if the extraction were too large — it did not."
    },
    {
      "command": "npm run lint (tsc --noEmit && eslint src)",
      "result": "pass",
      "evidence": "Exit 0, no output."
    },
    {
      "command": "wc -l src/App.tsx src/hooks/useInventoryState.ts",
      "result": "pass",
      "evidence": "App.tsx 217, useInventoryState.ts 163 — both under the 250-line cap, down from a single 306-line file."
    },
    {
      "command": "bash ~/utils/scripts/check_file_length.sh --all",
      "result": "pass",
      "evidence": "No kcd2_dice_solver file appears in the over-cap listing at all. This is the fourth and last of the four files the series targeted."
    },
    {
      "command": "grep -rn 'eslint-disable|ts-ignore|ts-expect-error' kcd2_dice_solver/src/ ; git diff --stat vite.config.ts",
      "result": "pass",
      "evidence": "No hits and an empty diff: zero suppressions across the whole series, and the coverage thresholds and exclude list were never touched."
    },
    {
      "command": "pre-commit run --files <changed files>",
      "result": "pass",
      "evidence": "All configured hooks pass on the touched files."
    }
  ],
  "risks": [
    "The hook has no dedicated test file; its coverage comes entirely from the four App test files exercising it through the component. Coverage is 100% including branches, but a future change to the hook could be under-tested in a way a direct hook test would catch. Adding one was deliberately left out of scope because it would change the 288-test baseline this split is verified against.",
    "App.tsx and the hook now share a wide interface — eleven fields. That is the cost of extracting state rather than markup from a component with no prop-drilled subtree; a presentational split would have needed comparable prop plumbing.",
    "kcd2_dice_solver has no CI workflow and is not covered by pre-commit, so the vitest and lint runs above are the only gate on this change."
  ],
  "rollback": [
    "git revert the commit — one edited file and one added file, with no data, config or schema migration.",
    "After rollback, re-run npx vitest run --coverage and npm run lint; expect 288 tests passing at 100% coverage (totals returning to 863/375/181/810) and lint exit 0."
  ]
}