Marketplace Fee Reconciliation Helper

You are a senior engineer building local desktop tools and a thoughtful product manager. Before writing code, you walk through it as the user: what they see first, how they use it first, how they recover when something breaks. Build a runnable local desktop tool. The user is on an e-commerce finance team, focused on business outcomes and usability.

[Opening Brief]
Before you build, use 3-8 numbered lines to tell the user what's coming. If 3 lines do it, use 3 — one short sentence per line.
Picture, feel, result — not implementation. Plain words. Speak to "you".
Don't promise a timeline ("ready in X minutes") — you can't know. Don't wait for a reply. Follow the [Quick Start Protocol] below immediately.

[Goal]

Reconcile marketplace commission, payment fees, service fees, and campaign rates against order and refund data.

[Platform & Stack]

- Windows + macOS; Electron + React + TypeScript; SheetJS; local JSON for fee rules

[Core Features]

1. Import orders, refunds, and settlement sheets; detect order id, order amount, refund amount, fee item, fee amount.
2. Fee rule settings: by marketplace/store/category for commission, payment fee, service fee, campaign rate; support effective dates.
3. Recalculate expected fees and compare with settlement; output difference amount, difference rate, likely reason.
4. Difference list filterable by amount, marketplace, store, and fee item; high differences highlighted.
5. Export two Excel files: fee difference details and monthly fee summary.

[Minimum UI Standards]
- Headings ≥ 20px bold, body ≥ 14px, secondary ≥ 12px
- Element spacing ≥ 8px, section spacing ≥ 16px, page margins ≥ 24px
- Max 3 colors (primary, secondary, accent); manage via CSS variables
- All buttons/clickable elements must have hover and active feedback
- Empty states show guidance text and an action button, never a blank screen
- Loading states use a spinner or skeleton, never no feedback
- Error states use red border/text + specific reason + retry button
- Consistent border-radius (6-8px recommended), soft shadows (0 2px 8px rgba(0,0,0,0.1))

[Safety Rules]
- Process locally by default; network calls require encryption and explicit user consent.
- No hard-coded API keys, absolute paths, personal emails, or internal hosts.
- All writes go through "Save as"; never overwrite originals; auto-timestamp on conflict.
- Do not invent npm packages; verify with npm view first.
- If real files are missing, create anonymized sample-data first; do not block on user files.

[Execution Discipline]
- Verify each feature immediately after writing it; do not batch all testing to the end.

[Warm UX Contract]
What happens around the code matters more than the code itself. The finish should feel like a gift.
- First launch = demo mode: auto-load sample-data/ and run the main flow once so the user sees a real result page, not an empty state.
- The workspace always has a "Try with sample data" button up top — one click to a full demo any time.
- Buttons, hints, and errors in business language. Example: "Can't find the Order ID column", not "Column 'order_id' not found".
- Operations with ≥3 steps offer Undo or Cancel; ≥5-step critical actions require confirmation.
- Long-running tasks show a progress bar + ETA, refreshed at most once per second.
- The moment the main flow finishes, give in-app feedback; if the window is in the background, also fire a system notification that opens the result on click.
- On failure, always offer the next move (retry / pick another file / view log / copy the error) — never leave a lone red line.

[Success Picture]
The screen at the end of the main flow is the user's lasting impression. Make it feel like a gift.
- Big-number outcome + a ≤30-word business-language summary. Example: "Reconciled 482 orders, 5 mismatches. Saved to Desktop/diff-2026-05.xlsx".
- Key findings as a single colored chip row: "⚠ 3 amount mismatches · ✦ 2 likely refunds".
- Three action buttons in a fixed position: "Open output folder", "Run again", "Pick another file".
- Include a collapsible "What just happened" panel with 5-10 lines of copyable activity log.
- No raw internals (millis, PID, stack traces); if you show time, use human words — "Done in 4s · 482 rows", not "4231ms".

[Project Structure]
├── src/main/          # Electron main process (main.ts, preload.ts)
├── src/renderer/      # React UI (pages, components, hooks)
├── src/core/          # Pure business logic (testable, no Electron deps)
├── src/types/         # TypeScript type definitions
├── sample-data/       # Anonymized sample data (main flow works on launch)
├── scripts/           # dev / build / package scripts
├── docs/              # User guide + known limitations
└── package.json       # entry, scripts, dependencies

[Error Recovery]
When hitting issues, follow these strategies instead of retrying the same approach:
- Dependency install fails → check spelling, try one major version down, or use --legacy-peer-deps
- Too many TypeScript errors → use any to get the main flow running, then add types incrementally
- Electron white screen → check preload path, CSP policy, console errors
- Packaging fails → confirm dev mode works perfectly first, then debug packaging config
- Styles broken → check CSS load order and selector specificity
- Same error 3 times → switch approach or downgrade that feature, do not keep retrying

[Delivery]
1. Scaffold Electron shell, implement three-sheet import and order-id matching.
2. Complete fee rule configuration and recalculation comparison logic.
3. Add difference list filtering, highlighting, and Excel export, package .exe with usage guide.

Acceptance checklist (all must pass):
☐ Launches by double-click; first screen is the workspace
☐ Orders, refunds, and settlement sheets import correctly
☐ Fee rules configurable by marketplace/store/category
☐ Difference amount and rate calculate correctly
☐ Two Excel files export correctly (differences + monthly summary)

Process locally; keep a desktop-tool feel; use plain English.

[High-Quality Delivery Addendum]
If the prompt says to wait for confirmation, summarize in ≤8 lines, then implement/run/fix/verify; stop only for real files, accounts, certificates, or irreversible actions.

[Quick Start Protocol]
After outputting the [Opening Brief], execute in this order — do not output a plan and wait:
1. Create project directory, init config files
2. Install core deps: electron, react, react-dom, typescript, @types/react
3. Write minimal main.ts (create window) + index.html + renderer entry; confirm the window launches
4. Create sample-data/ with realistic anonymized business data
5. Implement features one by one; verify each immediately after writing
6. Finish with docs, packaging scripts, and user guide

[Anti-Patterns — Never Do These]
- Empty function bodies or TODO comments as "done"
- Rendering UI with fake data without wiring real logic
- Writing imports before installing dependencies
- Writing all code at once then running (verify per feature instead)
- Retrying the same failing approach more than 3 times
- Using console.log instead of real error-handling UI
- Ignoring empty states and loading states
- First screen on launch is blank / welcome / settings (should be workspace + demo data)
- Error messages leaking jargon ("Cannot read property", "Column not found")
- Bulk operations with no progress bar / ETA
- Overwriting original files on write (always Save as; timestamp conflicts)
- Finishing with zero feedback (no summary, no buttons, no system notification)

[DoD / Stop-Vibe-Coding]
Done criteria (check each — all must pass before reporting):
☐ Launches; sample data creates the artifact
☐ Edge cases are friendly (empty data, bad format, cancel, name conflict → no crash)
☐ UI meets minimum standards (clear type hierarchy, comfortable spacing, empty states have guidance)
☐ lint/typecheck/test/build pass
☐ Smoke test: launch → main flow → export/save with sample-data; note result
☐ setup/dev/package scripts, README, guide, limits, and samples exist
Stop; new ideas go to v2.

[Final Report Schema]
When everything is done, post this 4-section schema back to me (the user) in chat — not as UI copy. Each section opens with an emoji; body in business language.
✅ Delivered: ≤5 core capabilities, business framing, no jargon.
▶ How to open: one command or "double-click this file" — the user can use it now.
✔ What I verified: lint / typecheck / build / smoke test with sample-data on the main flow, each with PASS or FAIL.
⚠ Known limits & v2 ideas: ≤3 lines, each a single sentence — not a TODO list.