amazon-reports

name	amazon-reports
description	Amazon platform only. MUST load BEFORE taking any action when the task involves Amazon Seller Central report pages (Tax Document Library, Business Reports, Fulfillment, Payments CSV, Advertising Reports, etc.). Contains URLs, hover navigation, CSV structures, and wait times.
requires	["amazon-shared"]
review	{"criteria":"- The requested report(s) were ACTUALLY exported: the file exists,\n is non-empty, and has rows > 0 — OR the page explicitly shows \"No\n Data Available\" (a valid empty result, not a missing/failed pull).\n- Scope matches the ask: every requested country / the requested date\n window is covered, not a partial pull.\n","verify_by":"Open the newest downloaded file in ~/.vibe-seller/downloads/<slug>/\nand confirm the header columns + a non-zero row count; open the\nreport-history page and confirm the \"Date Range Covered\" and scope\nmatch the request. Do not accept a claimed export without the file.\n"}

Amazon Seller Central — Reports Export Guide

PREREQUISITE: Read ../amazon-shared/SKILL.md for marketplace TLD map, the hamburger-menu hover pattern (referenced below), sign-in / Ziniao / OTP handling, and the capture rule.

This skill documents how to navigate to and export every report type available in Amazon Seller Central and Amazon Advertising, including CSV/PDF structures for building analysis scripts.

Critical: Hamburger Menu Navigation

Many reports are accessed via the hamburger menu (top-left corner). This menu uses a hover-to-reveal pattern — you must dispatch a mouseover, not a click, on a category to reveal its submenu items. (See ../amazon-shared/SKILL.md § 4 for the canonical version.)

browser-use <<'PY'
# 1. Open the hamburger menu (opens the full sidebar overlay).
#    The button is a role=button inside the navigation-hamburger-menu
#    shadow DOM — reach it through its shadow host.
js("document.querySelector('navigation-hamburger-menu').shadowRoot.querySelector('[role=button]').click()")
wait_for_load()

# 2. Reveal the submenu by dispatching mouseover on the "Reports"
#    category (aria-expanded flips false -> true; flyout appears).
js("""
var cat = [...document.querySelectorAll('[aria-expanded=false]')]
  .find(e => e.textContent.trim() === 'Reports');
cat.dispatchEvent(new MouseEvent('mouseover', {bubbles: true}));
""")
print(page_info())

# 3. Click the submenu item once it renders (e.g. "Tax Document Library").
js("""
[...document.querySelectorAll('a, [role=button]')]
  .find(e => e.textContent.trim() === 'Tax Document Library').click();
""")
PY

Do NOT click the category name — clicking navigates away instead of revealing the submenu. Always dispatch mouseover first, then print(page_info()) to see the submenu items, then click the target. If a stable selector isn't obvious, fall back to capture_screenshot() then click_at_xy(x, y).

General Rule: Download Existing Reports First

Before generating any report, always check if a matching report already exists. All report pages maintain a download history table showing previously generated reports with their date ranges and Download buttons.

This applies to ALL report types:

• Fulfillment reports (/reportcentral/): Click the report in the left sidebar → "Download" tab → history table with columns: Report Type | Date Range Covered | Date Requested | File Format | Report Status (Download button). If a row matches the requested date range, click Download directly.

• Advertising reports (Unified Reporting, advertising.amazon.{tld}/reporting): the report list shows saved/generated reports with a 报告状态 column; click into a 已完成 (Completed) row for the Download link. If a completed run matches the date range, download it. (The old /reports console is retiring 2026-12-31 — see §6.)

• Payments reports (/payments/reports-repository): Check the report list for existing reports covering the target period.

Only request/generate a new report if no existing one covers the needed date range.

An empty export is a SUCCESS; "N/A" must be proven on the page

Two rules that together stop the biggest false-failure pattern (a store was marked FAILED for reports it "doesn't have" — decided from metadata, never checked on the page):

1. Store metadata does NOT tell you which reports apply. The store's platforms / countries (e.g. {"amazon": ["SA"]}) only say which marketplaces it sells on. They encode nothing about FBA enrollment or whether an Advertising account exists. Never conclude "no FBA → skip storage/returns" or "no Ads → skip the ad report" from metadata. You must open the actual page and let it tell you.

2. A report that exports with zero data rows is DOWNLOADED, not missing. Request each report on its page:

• Ads report: warm the ad console first (via Campaign Manager in the Seller Central menu, or the choose-account?destination=/reporting flow in §6.1) — a cold direct advertising.amazon.{tld} URL gives a "Sign in / Register" marketing page. Once warm, build the report in Unified Reporting (§6) — Amazon generates it even with zero campaigns (headers, 0 rows) → download that empty file, it's a completed deliverable. The ad report is genuinely N/A only if, after warming up, the store shows an advertiser onboarding / registration flow (no advertiser account). A marketing landing reached by direct URL proves nothing — it just means you weren't SSO'd; it is NOT evidence of "no Ads".
• FBA storage / returns: open the report page and request the month. An empty result / "No Data Available" for a valid request still means you asked correctly — download whatever file is produced. Only if the page explicitly says the store is not enrolled in FBA is it N/A.

When to use each task outcome:

• downloaded (incl. empty 0-row exports) → deliverable met.
• N/A — only when the page proved the capability is absent (advertiser-registration landing; explicit not-enrolled-in-FBA). Record in vibe_seller_set_task_result; do NOT vibe_seller_set_task_error.
• pending-Amazon-latency (e.g. Monthly Storage Fees not yet published, see §2) → vibe_seller_set_task_result, not vibe_seller_set_task_error.
• failed → vibe_seller_set_task_error only for a report the page shows you should be able to get but couldn't (dead button, 0-byte download, error page).

If the only gaps are proven-N/A or latency-pending, the task COMPLETES. State per report which of {downloaded, N/A-proven-on-page, pending-latency, failed} it is, so the outcome is unambiguous — and never downgrade "I didn't check" into "N/A".

Clicking a Download button (Amazon kat-* shadow DOM)

Report tables wrap the Download control as kat-table-row → kat-table-cell → kat-button, and the real clickable <button> lives inside kat-button.shadowRoot. A plain document.querySelector('kat-table-cell button') (or any [data-testid=...] / .download-btn guess) returns null — CSS selectors do not cross shadow boundaries, and [...document.querySelectorAll('button')] never sees a button that lives inside a shadow root. So those selectors silently no-op and you waste turns. Reach the inner button through the shadow host and .click() it. (This .click() works for table Download buttons; the left-sidebar kat-buttons are the exception — those need a coordinate click.)

Use this one snippet on every report page. Match the target row by a substring of its date range, then click its Download control:

browser-use <<'PY'
print(js(r"""
var rowMatch = '01/06/26';   // substring of the TARGET row's date range
var rows = document.querySelectorAll('kat-table-row, tr');
for (var r of rows) {
  if (!(r.textContent || '').includes(rowMatch)) continue;
  for (var h of r.querySelectorAll('kat-button, kat-link, a, button')) {
    var real = h.shadowRoot ? h.shadowRoot.querySelector('button, a') : h;
    var label = (h.textContent||'') + ' ' + ((h.getAttribute && h.getAttribute('label')) || '');
    if (real && (h.tagName === 'KAT-BUTTON' || /download|\.csv/i.test(label))) {
      real.click();
      return 'clicked download for row ~ ' + rowMatch;
    }
  }
}
return 'NO download control for ~' + rowMatch + ' (check the date substring / row exists)';
"""))
import time; time.sleep(3)   # let the download start
PY

On "NO download control", fall back to print(page_info()) (it often lists a plain Download link, clickable by index) or capture_screenshot()

• coordinate click — but try the snippet first; it one-shots the common kat-button shadow case. The same host-then-inner-button walk applies to any kat-button (e.g. "Request Report", "Download CSV") — match on its label/text instead of the date substring.

Setting a custom date range (works across every Seller Central variant)

Date controls differ per marketplace and per report page — and Amazon changes them over time. You will meet native <select> presets, kat-dropdown presets, plain <input> date fields, kat-date-picker / kat-date-range-picker shadow widgets, and month/year dropdowns — the same report can render differently on .sa vs .ae vs .com. Do not assume a widget. Follow the same three-step loop everywhere: observe → interact the human way → verify the outcome. The outcome check is DOM-independent and is what makes this robust.

Step 1 — Observe. capture_screenshot() and glance at the DOM (document.querySelector('kat-date-range-picker')? a native <select>? plain input[placeholder]?). Identify the preset control and, once a custom range is chosen, the two date fields.

Step 2 — Interact the way a human would for THAT widget. Prefer the interaction that fires the component's real handlers:

• Preset → "Exact dates": a native <select> takes a value-set (el.value='…' + change); a kat-dropdown does not — click_at_xy to expand, then click_at_xy the "Exact dates" row.
• The two date fields: the reliable path on every variant is the calendar popup with real clicks — click the field's calendar icon, click_at_xy the ‹ / › month arrows to the target month, then click_at_xy the day cell. (Locate cells from the screenshot — they aren't reliable via page_info()/JS rects.) Plain <input> fields on some marketplaces also accept fill_input(selector, "<date>") — try it if they truly are light-DOM inputs, but only trust it after Step 3.

⚠️ On kat-* date widgets, programmatic value-setting is display-only and silently ignored on submit — live-verified on AE FBA Customer Returns (2026-07-07): setAttribute('start-value'), the startValue/endValue property, the nested shadow <input>.value + input/change events, and CDP Input.insertText all updated the visible field (and even rp.startValue) yet the report generated for today. dropdown.selectOption('-1') likewise sets .value without emitting the reveal event. The component's submit reads an internal model only real user input updates — so calendar clicks, not scripts.

Step 3 — Verify the outcome (the universal acceptance test). This is the check that survives every DOM difference:

1. A quick sanity read after a calendar click — js("var rp=document.querySelector('kat-date-range-picker'); return rp?rp.startValue+' -> '+rp.endValue:'n/a'") — is fine as a first look, but it is not authoritative: a programmatic set can leave startValue/endValue showing your target without the value actually committing (see the warning above), so a match here does not guarantee the report will use it.
2. The authoritative check: request the report, then CONFIRM the new history row's "Date Range Covered" is your target month, not today. A today-dated (or wrong) row means the date never committed — go back to Step 2 and use real clicks; do not download the wrong-range file.

Because the calendar fills the field itself, date-format variance (DD/MM/YYYY / MM/DD/YYYY / YYYY/M/D) is irrelevant — never hand- type the format. Month/year dropdowns (Payments Reports Repository, Monthly Storage Fees) are kat-dropdowns → select by real click, never selectOption(); month index has been seen 0-indexed (Jan=0) on some pages — confirm against the visible label, don't hardcode it.

Report-status polling — keep each sleep short. After requesting a report, poll for the Download button; do not time.sleep(120) in one call — a single long sleep exceeds the browser-use tool timeout and the turn is killed mid-wait. Loop with short sleeps and re-check instead:

browser-use <<'PY'
import time
for _ in range(6):          # ~6 × 30s ≈ 3 min, each well under the tool timeout
    time.sleep(30)
    info = page_info()
    if 'Download' in info or 'Ready' in info:
        break
print(info)
PY

Report Types Overview

A. Seller Central Reports (via hamburger menu → Reports)

#	Report	Direct URL Path	Export Type	Wait Time
1	Business Reports	/business-reports	CSV download	Instant
2	Custom Analytics	(new feature)	Varies	Varies
3	Fulfillment	/reportcentral/WelcomePage	Request → Download	1-5 min
4	Advertising Reports	Redirects to Ad console	CSV download	Hours/Days
5	Return Reports	/reportcentral/WelcomePage	Request → Download	1-5 min
6	Custom Reports	/listing/reports	Request → Download	1-5 min
7	Inventory Reports	/inventory-reports	Request → Download	1-5 min
8	Tax Document Library	/tax/seller-fee-invoices	View (PDF)	Instant
9	Manage Taxes	/tax/tax-settings	Settings page	N/A
10	Selling Economics & Fees	/selling-economics-and-fees	Dashboard	Varies

B. Payments Reports

Report	Direct URL Path	Export Type	Wait Time
Reports Repository	/payments/reports-repository	Request → CSV	1-3 min

C. Advertising Reports (separate console)

Report	URL Pattern	Export Type	Wait Time
Ad reports — Unified Reporting	advertising.amazon.{tld}/reporting	CSV download	few min – 30+ min

---

1. Business Reports

Navigation: hamburger menu → hover Reports → Business Reports Direct URL: https://sellercentral.amazon.{tld}/business-reports Lands on: #/dashboard (Sales Dashboard)

Sub-sections (left sidebar)

• Dashboards → Sales Dashboard
• By Date → Sales and Traffic, Detail Page Sales and Traffic, Seller Performance
• By ASIN → Detail Page Sales and Traffic, By Parent Item, By Child Item

Export Flow

browser-use <<'PY'
new_tab("https://sellercentral.amazon.{tld}/business-reports")
wait_for_load()
print(page_info())   # locate the "Download" button (inside a kat-table-cell shadow DOM)
PY
# Trigger the CSV download with the canonical shadow-DOM download snippet
# (see "Clicking a Download button" above). The Business Reports view has a
# single Download button, so match on its label rather than a date row:
#   real = h.shadowRoot ? h.shadowRoot.querySelector('button,a') : h  → real.click()

Wait Time

Instant — page data is pre-loaded, Download button exports current view.

---

2. Fulfillment Reports

Navigation: hamburger menu → hover Reports → Fulfillment Direct URL: https://sellercentral.amazon.{tld}/reportcentral/WelcomePage

Sub-reports (left sidebar, elements have id=report-nav-link-url)

Category	Reports
Inventory	Inventory ledger report, Stranded Inventory, Reserved Inventory, Amazon Fulfilled Inventory, Manage FBA Inventory
Shipping	Inbound Performance, Outlet Deals
Orders	Amazon Fulfilled Shipments, Amazon Fulfilled Shipments – Tax Invoicing, All Orders
Fees	Fee Preview, Monthly Storage Fees, Aged Inventory Surcharge report
Returns	Reimbursements, FBA customer returns, Replacements, Removal Order Detail, Removal Shipment Detail
Other	EPR Category Reports

Export Flow

browser-use <<'PY'
new_tab("https://sellercentral.amazon.{tld}/reportcentral/WelcomePage")
wait_for_load()
print(page_info())   # sidebar report links carry id=report-nav-link-url
PY
# Click the sidebar link and the "Request Report" button by their
# page_info() index (they are kat-* controls; a raw querySelector guess
# no-ops). For any kat-button, use the host→inner-button walk from
# "Clicking a Download button" above, matching on its label text.
# Set the date range first via "Setting a custom date range" above.
# Wait 1-5 minutes, then download with the canonical download snippet,
# matching the history row by its date-range substring.

Wait Time

1-5 minutes for most reports. Refresh the page to check status. Poll with short sleeps (never one time.sleep(120) — see the polling loop in "Setting a custom date range" above).

Monthly Storage Fees — publication latency

Observed behavior: requesting Monthly Storage Fees for the previous month early in the new month returns No Data Available on the detail page even though the request itself is accepted. The same month's range re-requested about a week later returns a Download button within ~30s. Older months (M-2, M-3) request successfully right away.

So Storage Fees for month M is not available on day 7 of M+1 but is available by day 13-14 of M+1. The exact day Amazon publishes is not documented on the page; the page header for this report says only "Report data 24 hours old" (about freshness of already-published values, not about when they appear).

How to handle a No Data Available Storage Fees response when the requested month is the previous calendar month:

1. Don't keep retrying within the same task — Amazon won't surface the data sooner because of repeat requests.
2. Produce every other expected file (Transaction CSV, FBA Customer Returns CSV, Sponsored Products advertised-product XLSX) — those publish on a different cadence and ARE available today.
3. A scheduled re-run after roughly day 13-14 of the following month will capture the storage CSV.

Task-status decision for this scenario — MANDATORY exception to the system-prompt's "MUST call vibe_seller_set_task_error when primary objective could not be fully completed" rule:

If the ONLY missing deliverable is storage.csv (any country) and the cause is this Amazon publication-latency window (today is in month M+1, Amazon's detail page says No Data Available), then:

• Call vibe_seller_set_task_result(<full report including a "Storage CSV pending Amazon publication" note>) with the partial summary.
• DO NOT call vibe_seller_set_task_error. The task lands COMPLETED — the next scheduled run picks up storage when Amazon publishes it. Calling set_task_error here causes the task to land FAILED for what is a known, expected, upstream-paced data state, and that triggers false "task failed" alerts downstream.

This is a NARROW exception. It applies only when: (a) every other deliverable for the task ran successfully, AND (b) the only gap is storage.csv, AND (c) the agent saw No Data Available on the report-history detail page for that exact month.

If ANY other deliverable also failed, fall back to the normal "call BOTH set_task_result and set_task_error" pattern from the system prompt — the storage-latency exception does not absorb unrelated failures.

If No Data Available shows for an older month (≥ ~30 days old), that's unexpected — investigate normally; do not wave away, and do call set_task_error.

FBA Customer Returns — Exact dates flow (MANDATORY for month ranges)

Direct URL: https://sellercentral.amazon.{tld}/reportcentral/CUSTOMER_RETURNS/1

The date control is NOT a From/To pair like other reports. It is a preset dropdown (initial title last day (yesterday)) whose options are last 1/3/7/14/30/90/180 days — plus an Exact dates option (kat-option value=-1) at the bottom. Presets are anchored to today, so "last 30 days" does NOT equal a calendar month: on June 3 it covers May 4 → June 3, silently dropping May 1–3 and mixing in June rows. Never use a preset when the task asks for a calendar month.

Correct flow:

Select Exact dates FIRST — the .csv download control only appears after an exact range is chosen (with a preset selected, only a TSV button shows).

Open the page, then set the range with the real-click calendar method in "Setting a custom date range" above — screenshot, click the preset dropdown, click "Exact dates", then set start/end from the calendar popups. Do not try selectOption('-1') or any programmatic date-set: live-verified that they leave the request on today's date.

browser-use <<'PY'
new_tab("https://sellercentral.amazon.{tld}/reportcentral/CUSTOMER_RETURNS/1")
wait_for_load()
import time; time.sleep(3)
capture_screenshot()   # locate the preset dropdown + (after Exact dates) the calendar icons
PY

Finally request and download — the "Request .csv" control is a kat-button, so use the host→inner-button walk from "Clicking a Download button". Then confirm the new history row's date range is the month you asked for, not today, before downloading (a today-dated row = the date never committed; redo the calendar clicks). The .csv control only appears once an exact range is set; a preset shows only TSV.

UTC edge rows are normal: the date filter is marketplace-local, but the CSV's return-date column is UTC. A row dated the last day of the previous month at ~21:00–23:59 UTC is the 1st of the requested month in local time — keep it, don't re-request.

The download lands as a numeric report id (e.g. 51901234567.csv) in the store downloads dir; rename to return.csv when the task asks for that exact name.

---

3. Tax Document Library

Navigation: hamburger menu → hover Reports → Tax Document Library Direct URL: https://sellercentral.amazon.{tld}/tax/seller-fee-invoices

Page Structure

• Tab: Seller Fee Invoices (default and only tab)
• Displays most recent 1,000 invoices. Click "Load More" at bottom for older.
• Invoices grouped by month (newest first)

Table Columns

Invoice Type | Invoice File Type | Invoice Number | Payer Name | Payer Registration |
Supplier Name | Supplier Registration | Marketplace | Start Date | End Date | Action

Invoice Types

Invoice Type	Description
Fulfillment by Amazon Tax Invoice	FBA service fees invoice
Merchant VAT Credit Note	VAT credit adjustment
Merchant VAT Invoice	VAT invoice for seller fees
Product Ads VAT Invoice	Advertising VAT invoice

Invoice Number Format

{CC}-{ENTITY}-INV-{YEAR}-{SEQ} (invoices) or {CC}-{ENTITY}-CN-{YEAR}-{SEQ} (credit notes)

Where {CC} = country code (e.g. US), {ENTITY} = Amazon entity code, {SEQ} = sequence number.

Export Flow

browser-use <<'PY'
new_tab("https://sellercentral.amazon.{tld}/tax/seller-fee-invoices")
wait_for_load()
print(page_info())   # locate View buttons (id=view_invoice_button-announce)

# Each "View" button has a value=urn:alx:ver:{UUID} attribute.
# Clicking triggers a form POST that downloads a PDF. To pick a
# specific invoice, match its row; the first View button is:
js("document.querySelector('#view_invoice_button-announce').click()")

# For a specific month, scan page_info() output for date strings
# e.g. "Mar 01" or "Mar 31" for March invoices.

# For older invoices beyond the most recent 1,000, scroll to the
# bottom and click "Load More":
js("window.scrollTo(0, document.body.scrollHeight)")
print(page_info())   # locate "Load More"
js("[...document.querySelectorAll('button, a')].find(e => e.textContent.trim() === 'Load More').click()")
PY

Wait Time

Instant — invoices are already generated, View triggers immediate PDF download.

PDF Content (for analysis)

Each PDF is a formal tax invoice containing: invoice number, dates, seller and supplier VAT details, line items with amounts, VAT calculations, totals. These are individual documents — not bulk-exportable as CSV.

---

4. Custom Reports (Listing Reports)

Navigation: hamburger menu → hover Reports → Custom Reports Direct URL: https://sellercentral.amazon.{tld}/listing/reports

Report Types (dropdown values)

• Active Listings Report
• All Listings Report

Table Columns

Report Type | Batch ID (timestamp) | Report Status | Action (Download)

Export Flow

browser-use <<'PY'
new_tab("https://sellercentral.amazon.{tld}/listing/reports")
wait_for_load()
print(page_info())   # locate "Select Report Type" dropdown and "Request Report"

# Select the report type. If it is a kat-dropdown, use its native API
# (see "Setting a custom date range" above):
#   d.toggleExpanded(true); d.selectOption('<value>')
# A plain [data-testid=...] / <select> querySelector no-ops on a kat-*
# control. Then click "Request Report" via the host→inner-button walk
# from "Clicking a Download button", matching its label text.
PY
# Wait 1-5 minutes (poll with short sleeps — see the polling loop above),
# then download the new row with the canonical download snippet from
# "Clicking a Download button", matching the row by its date-range /
# request-time substring.

Wait Time

1-5 minutes. Refresh page to check if Report Status = ready.

Multi-country downloads collide on filename — rename each immediately

The All Listings / Active Listings Report downloads with a fixed, date-based filename (e.g. All+Listings+Report_MM-DD-YYYY.txt) that is identical across marketplaces. If you download the report for two countries in a row (e.g. SA then AE), the second silently overwrites the first in the download dir — no browser prompt. Two consequences:

• Rename right after each download, before triggering the next. Copy the freshly-downloaded file to a country-suffixed name (All_Listings_<CC>_<YYYY-MM-DD>.txt) the moment it lands, then request the next country.
• The All Listings TSV is account-level (byte-identical across a unified account's marketplace subdomains), so a per-country re-download often yields the same rows anyway — verify content rather than assuming each country's file differs.

---

5. Payments Reports Repository

Navigation: hamburger menu → hover Payments → Payments, then tab "Reports Repository". Or navigate directly. Direct URL: https://sellercentral.amazon.{tld}/payments/reports-repository

Page Tabs

Tab	Description
Statement View	Settlement period summary: Net Proceeds, Sales, Refunds, Expenses, Balance
Transaction View	Individual transactions within a settlement
All Statements	Historical settlement statements
Disbursements	Bank disbursement details
Advertising Invoice History	Ad billing invoices
Reports Repository	Custom date-range CSV reports (selected by default via direct URL)

Report Type Dropdown Values

Value	Label	Description
SELLER_TRANSACTION_DATE_RANGE	Transaction	Per-order/per-event transaction details
SELLER_SUMMARY_DATE_RANGE	Summary	Aggregated summary for the period

Export Flow

browser-use <<'PY'
new_tab("https://sellercentral.amazon.{tld}/payments/reports-repository")
wait_for_load()
print(page_info())

# 1. Report Type dropdown (id=katal-id-* with title=Transaction)
js("document.querySelector('[title=Transaction]').click()")
print(page_info())   # locate kat-option elements
js("[...document.querySelectorAll('kat-option')].find(o => o.textContent.trim() === 'Transaction').click()")  # Transaction or Summary

# 2. Date Range — set start and end dates with the REAL-CLICK calendar
#    method in "Setting a custom date range" above (screenshot → click
#    the field's calendar icon → month arrows → click the day cell).
#    Programmatic date-sets do NOT commit. If the page offers a "Month"
#    radio, click it and pick month + year from the kat-dropdowns by
#    real clicks (month observed 0-indexed). Verify the picker shows your
#    range before requesting.

# 3. Request Report — kat-button; use the host→inner-button walk from
#    "Clicking a Download button", matching its label.
PY
# 4. Wait ~1-3 minutes (poll with short sleeps — see the polling loop
#    above), then scroll down and download the "Ready" row (check_circle
#    icon) with the canonical download snippet, matching on the
#    "Download CSV" label. A raw querySelectorAll('button') never sees
#    the button inside kat-button.shadowRoot.
browser-use <<'PY'
js("window.scrollTo(0, document.body.scrollHeight)")
print(page_info())
PY

Wait Time

1-3 minutes for report generation. The table row shows a status icon:

• Spinner = generating
• Green check (kat-icon name=check_circle) + "Ready" = downloadable

Download Mechanism

The "Download CSV" button triggers an XHR GET to:

/payments/reports/api/download-report?reportId={UUID}

This returns the CSV file. The browser handles it as a download.

Transaction Report CSV Structure

The Transaction report has 7 header lines (definitions) followed by a 22-column CSV with ~17,000+ rows per quarter.

Header lines (skip first 7 lines when parsing):
  Line 1: "Includes Amazon Marketplace, Fulfillment by Amazon (FBA), and Amazon Webstore transactions"
  Line 2: "All amounts in {currency}, unless specified"  (e.g. AED, USD, EUR)
  Lines 3-7: Column definitions
  Line 8: Column headers
  Line 9+: Data rows

Columns (22 total)

"date/time","settlement id","type","order id","sku","description","quantity",
"marketplace","fulfillment","order city","order state","order postal",
"product sales","shipping credits","gift wrap credits","promotional rebates",
"sales tax collected","Marketplace Facilitator Tax","selling fees","fba fees",
"other transaction fees","other","total"

Column Details

Column	Type	Example	Description
date/time	datetime	15 Mar 2026 3:22:08 PM UTC	Transaction timestamp
settlement id	string	18472930156	Settlement period ID
type	enum	Order	See transaction types below
order id	string	123-1234567-1234567	Amazon order ID
sku	string	PROD-001	Your SKU
description	string	Product title	Full product name
quantity	int	2	Units in transaction
marketplace	string	amazon.<tld>	Marketplace domain
fulfillment	string	Amazon	Amazon (FBA) or Seller
order city	string	City name (may be in the local language)	Buyer's city
order state	string	State/region	Buyer's state
order postal	string	Postal code	May be empty
product sales	decimal	89.00	Product revenue
shipping credits	decimal	5.00	Shipping collected
gift wrap credits	decimal	0	Gift wrap revenue
promotional rebates	decimal	-4.45	Promotions (negative)
sales tax collected	decimal	4.45	Sales tax (5% VAT)
Marketplace Facilitator Tax	decimal	0	Amazon-collected tax
selling fees	decimal	-13.35	Referral fees (negative)
fba fees	decimal	-18.50	FBA fees (negative)
other transaction fees	decimal	0	Misc fees
other	decimal	0	Non-order amounts
total	decimal	62.15	Net amount

Transaction Types (column "type")

Type	Count (typical quarter)	Description
Order	~17,000	Product sale
Refund	~400	Customer refund
Adjustment	~120	Amazon adjustments
Service Fee	~28	Subscription/service charges
Transfer	~30	Bank disbursements
FBA Inventory Fee	~6	Storage/removal fees

Analysis Script Hint

import pandas as pd

df = pd.read_csv(
    'transaction_report.csv', skiprows=7, dtype={'order postal': str}
)
# Filter by SKU
sku_data = df[df['sku'] == 'YOUR-SKU']
# Profit per SKU
profit = df.groupby('sku')['total'].sum().sort_values()
# Monthly revenue
df['month'] = pd.to_datetime(
    df['date/time'], format='%d %b %Y %I:%M:%S %p %Z'
).dt.to_period('M')
monthly = df[df['type'] == 'Order'].groupby('month')['product sales'].sum()

---

6. Advertising Reports — Unified Reporting (统一报告 / new report center)

⚠️ Amazon is retiring the old ad-report consoles. The Sponsored Ads report console (advertising.amazon.{tld}/reports) and the Amazon DSP report console close on 2026-12-31 — after which all saved reports, scheduled reports, and their history are permanently deleted. New-report creation/editing in the old console is disabled from 2026-12-17 (existing reports go read-only). All new ad reporting goes through Unified Reporting (统一报告), the centralized report center documented below. Migration guide: advertising.amazon.com/help/GK3JZZE5QLTF8U8A.

Unified Reporting combines every ad product (Sponsored Products, Sponsored Brands, Sponsored Display, Sponsored TV, and Amazon DSP) and multiple accounts / marketplaces into one report, built from a shared library of standardized dimensions and metrics. One report can span products and countries that used to need several separate exports.

6.0 The two report systems (know which URL you're on)

System	URL	State
Unified Reporting (use this)	advertising.amazon.{tld}/reporting	Current. List + templates + custom builder.
Legacy Sponsored-ads console (retiring)	advertising.amazon.{tld}/reports	Read-only after 2026-12-17, deleted 2026-12-31. See §6.10.

The nav item 效果衡量和报告 → 报告 (Measurement & reporting → Reports) links to /reporting (unified). 搜索广告报告 (Search ads reports) still links to the old /reports console.

6.1 Navigation & warm-up

Same cold-session caveat as the old console: a cold direct hit to advertising.amazon.{tld}/reporting redirects to advertising.amazon.com/register (the marketing/onboarding page) — that is NOT evidence the account has no ads; it just means you are not SSO'd into the ad console yet. Warm the account first:

Warm-up via choose-account (reliable):

browser-use <<'PY'
new_tab("https://advertising.amazon.{tld}/choose-account?destination=/reporting")
wait_for_load()
# Click the advertising account button (its label = your ad-account name),
# then the marketplace button (e.g. 沙特阿拉伯 / 阿联酋 / United States).
js("[...document.querySelectorAll('button')].find(b => b.textContent.trim() === '<your-ad-account>').click()")
print(page_info())            # marketplace sub-buttons appear
js("[...document.querySelectorAll('button')].find(b => b.textContent.trim() === '沙特阿拉伯').click()")
wait_for_load()
print(page_info())            # → advertising.amazon.{tld}/reporting?entityId=<ENTITY>
PY

You land on advertising.amazon.{tld}/reporting?entityId=<ENTITY>. The entityId is per-account and per-marketplace — each marketplace of a multi-country account has its own entityId. Once SSO cookies are set in the profile, direct nav to .../reporting?entityId=<ENTITY> (and .../reporting/new?...) works warm for the rest of the session. You can also reach it in-console via the left nav 效果衡量和报告 → 报告. Do NOT hard-code an entityId from a past run — read it back from the URL after warm-up.

6.2 The report list page (/reporting)

Top of the page: report templates row (§6.3) + a 创建报告 (Create report) button for a custom report. Below: your saved/generated reports table with columns 报告名称 (Report name), 报告状态 (Report status), 创建时间 (Created), 上次修改时间 (Last modified), 上次运行时间 (Last run), 下次运行时间 (Next run), 账户 (Account), 报告提取频率 (Frequency). Empty state: "尚无 报告 — 首先选择模板或创建自定义报告".

A finished report is downloaded from here (§6.7). 报告状态 values: 排队/正在处理 (Queued/Processing — wait) → 已完成 (Completed — download available) → 失败 (Failed).

6.3 Templates (fastest path)

Three prebuilt templates sit at the top of /reporting, each with a 使用模板 (Use template) button. Clicking one opens the builder at /reporting/new?entityId=<ENTITY>&templateId=<id> with a sensible default column set pre-seeded:

Template (zh)	templateId	Equivalent old report
广告活动 (Campaign)	campaign	Campaign report
推广的商品 (Advertised product)	advertised_product	SP "Advertised product" (adProducts)
搜索词 (Search term)	searchterm	Search term report

Use 查看所有模板 (View all templates) for the full set.

Seeding is async and only happens via the template CLICK. After clicking 使用模板, the ~20+ default columns populate a second or two later — wait for them (poll for 删除/Delete buttons in the column list). Navigating directly to /reporting/new?...&templateId=advertised_product by URL loads the page but does NOT seed columns — you'd submit an empty report and hit the validation error in §6.4. Always enter the builder by clicking the template on /reporting.

browser-use <<'PY'
new_tab("https://advertising.amazon.{tld}/reporting?entityId=<ENTITY>")
wait_for_load()
# Click 使用模板 inside the 推广的商品 card (match the card that mentions
# 推广的商品 but not 搜索词):
js("""
  var btns=[...document.querySelectorAll('button,a')].filter(b=>b.textContent.trim()==='使用模板');
  for (var b of btns){ var p=b; for(var i=0;i<6&&p;i++){ p=p.parentElement;
    if(p && /推广的商品/.test(p.textContent) && !/搜索词/.test(p.textContent)){ b.click(); return; } } }
""")
wait_for_load()
import time; time.sleep(5)   # let default columns seed
print(js("[...document.querySelectorAll('button')].filter(b=>b.textContent.trim()==='删除').length"))
PY

6.4 Building / customizing a report (the builder)

The builder (/reporting/new, title 创建报告) is a single scrolling form with five cards. The UI is Amazon's storm-ui framework — every control carries a data-takt-id="storm-ui-*" and most have stable ids.

The 提交 (submit) button ignores synthetic .click(). Verified live: js("document.getElementById('urc_frb_create_report_button').click()") silently no-ops — submit with a real CDP click (click_at_xy(x,y) from the element's getBoundingClientRect). Most other controls (the 使用模板 template button, date-range presets, the picker 保存, column checkboxes) do respond to js("el.click()") in testing — but if any click appears to no-op, fall back to click_at_xy. Use fill_input(sel, text) / type_text(...) for text inputs.

A. Report name — input id="report-name-form:report-name-control-component-0" (≤256 chars, auto-filled 推广的商品 - <timestamp>). The field keeps its auto-filled value unless you clear it, so to set a clean name select-all and delete (or empty the value) first, then enter yours — don't assume a single fill_input/type_text call replaces the existing text. A unique, descriptive name makes the report easy to find later.

B. 筛选条件 (Filters) — four rows, each with an 添加/变更 (Add/Change) button:

Filter	Default	Change control (id=)	Notes
账户 (Account)	current account (1)	添加 button (data-takt-id=storm-ui-button)	Dialog: 添加 (single), 添加当前和未来账户 (a manager acct incl. future-linked), 添加所有显示的账户 (all shown) → 保存.
广告活动 (Campaigns)	所有广告活动 (all with traffic)	filter-form:campaigns-row-control-component-0-button	Pick specific campaigns by name/id.
国家/地区 (Country/Region)	所有国家/地区	filter-form:campaign-country-row-control-component-0-button	See §6.8 — restrict here for a per-marketplace file, or add the 国家/地区 dimension and split downstream.
广告产品 (Ad product)	所有广告产品	filter-form:ad-product-row-control-component-0-button	SP / SB / SD / STV / DSP. Legacy SB (no ad groups) and self-service streaming-TV campaigns are NOT supported yet.

C. 定制列 (Custom columns) — the heart of the builder. Two tabs: 尺寸 (Dimensions) and 指标 (Metrics), a category list on the left, a Search box (type any field name), and checkbox rows (label[data-takt-id=storm-ui-checkbox] — click to toggle). Chosen columns appear as reorderable "Sortable Item" rows each with a 删除 (Delete) button; drag to reorder.

Minimum to submit (enforced on 提交):

1. ≥1 time dimension (维度 → 时间),
2. ≥1 detail dimension (维度 → 详细程度: 广告主账户 / 广告活动 / 广告组 / 广告), and
3. ≥1 metric (指标).

Some fields grey out based on other choices (hover to see why, e.g. "选择主 IMDb 广告展示量时此列不可用"). Empty cells for a dimension that doesn't apply to a given ad product are normal — that's how one report holds several products.

Dimension categories (维度 → tabs on the left): 时间 (Time), 详细 程度 (Detail level), 商品 (Product), 广告库存来源 (Inventory source), 交易 (Deals), 技术 (Tech), 地理位置 (Geography), 受众 (Audience), 投放 方案 (Targeting), 转化 (Conversion). Notable fields:

Category	Key fields
时间	日期 (Date, → daily rows), 周, 月, 年份, 一周中的某一天, 小时, 日期范围 (Date range, → one start/end summary row)
详细程度	广告主账户 / …ID / …名称, 实体编号, 广告组合(名称/编号), 广告活动(名称/编号/状态/预算/竞价方案/起止日期…), 广告组(名称/编号/状态/类型…), 广告(名称/ID/格式/尺寸/语言)
商品	推广的商品 / 编号(=ASIN) / 名称 / 品牌 / 品类 / SKU / 站点; 达成转化的商品(编号/名称/品牌/品类); 商品相关度
广告库存来源	网站或应用程序, 广告位(名称/大小/分类)
技术	操作系统, 浏览器名称/版本, 设备类型, 环境
地理位置	国家/地区, 国家/地区代码, 区域, 指定市场区域(DMA), 城市, 邮政编码
受众	细分受众群(ID/名称/类型/来源/国家/状态), 频次组
投放方案	投放目标, 目标竞价, 投放类型, 投放状态, 投放方案, 匹配类型, 搜索词, 匹配的目标
转化	转化来源, 转化来源所有者, 转化定义/类型（亚马逊站外）, 归因类型
内容/直播 (STV/DSP)	内容类型/评级/标题/创作者, 直播活动(编号/名称/广告位/广告时段…)

Metric categories (指标 → tabs on the left): 投放 (Delivery), 成本 和费用 (Cost & fees), 触达 (Reach), 展示量份额 (Impression share), 互动 (Engagement), 亚马逊网站转化量 (On-Amazon conversions), 亚马逊设备转化量 (Device conversions), 亚马逊视频转化量 (Video conversions), 亚马逊站外 转化量 (Off-Amazon conversions), 已合并的转化量 (Combined conversions). Notable fields:

Category	Key fields
投放	展示量, 点击量, 点击率 (CTR), CPC, CPM, 可见展示量/可见率/vCPM, 无效展示/点击
成本和费用	成本 (= spend/花费), 总成本, 广告库存成本, 各类费用 (代理费/平台费/第三方费用 …)
触达	用户触达量, 平均展示频率, 家庭触达量, 频次分布 (1…10+)
展示量份额	展示量份额, 展示量份额排名, 搜索结果首页首位展示量份额
互动	视频四分位完成率/完播率/完整观看, 音频同类, 5 秒观看
亚马逊网站转化量	购买量, 销售额, 已售商品数量, 购买率, ROAS, 单次购买成本; 商品详情页浏览量, 加入购物车, 加入心愿单, 品牌搜索量, 品牌旗舰店浏览, 评论页访问, 订购省, 长期销售 — each in base / （推广） / （光环） / （品牌新客） and 归因于点击 / 归因于浏览 variants
亚马逊设备转化量	订阅注册, 应用商店启动/使用时长, Alexa 技能, Kindle KENP / KU 借阅
亚马逊视频转化量	加入播放列表, 下载视频播放, 租借, 预告片播放, 视频有效播放
亚马逊站外转化量	每 on-Amazon metric mirrored with the （亚马逊站外） qualifier + 潜在客户/结账/安装/页面浏览/搜索/注册/订阅
已合并的转化量	购买量/销售额/ROAS （已合并） = on-Amazon + off-Amazon combined

D. 报告期 (Report period) — control id="report-range-form:date-range-row-control-component-0" (default 最近 7 天). Opens a preset list (data-takt-id=storm-ui-date-range-picker-preset-selector) and a dual-calendar for custom ranges. Presets: 今天, 昨天, 最近 7 天, 本周, 上周, 最近 30 天, 本月, 上个月, 最近 90 天, 本季度, 上个 季度, 今年, 去年.

A preset auto-applies — do NOT click 保存 after one. Verified live: clicking a preset (e.g. 上个月) closes the picker and applies the range immediately; the 保存 (data-takt-id=storm-ui-date-picker-confirmation-control-save) button is only present for a custom calendar start→end selection — after a preset it isn't in the DOM, so a querySelector(...).click() on it returns null / throws. After clicking a preset, just verify the control label changed (e.g. reads 上个月). Click 保存 only when you picked custom calendar dates.

Lookback limits: daily/weekly up to 15 months; monthly/yearly up to 6 years. Max lookback varies per dimension/metric — some are shorter. Presets that exceed a selected dimension's limit are hidden.

# Set period = last calendar month (for the monthly export)
browser-use <<'PY'
js("document.getElementById('report-range-form:date-range-row-control-component-0').click()")
import time; time.sleep(2)
js("[...document.querySelectorAll('[data-takt-id=storm-ui-date-range-picker-preset-selector]')].find(e=>e.textContent.trim()==='上个月').click()")
# A preset auto-applies — do NOT click 保存 (it isn't in the DOM after a
# preset). Just confirm the control label now reads 上个月:
import time; time.sleep(1)
print(js("document.getElementById('report-range-form:date-range-row-control-component-0').textContent"))
PY

E. 投放设置 (Delivery) —

• 发送至 (Send to): optional email(s), input id=recipients-input-builder-input. Anyone with the emailed link can view the report (no Amazon account needed).
• 报告提取频率 (Frequency): dropdown (data-takt-id=storm-ui-dropdown-trigger-button; items storm-ui-dropdown-item) = 一次 (Once, default) / 每日 / 每周 / 每月. Recurring → pick weekday (weekly) or day-of-month (monthly), an end date, and a time + timezone.
• 文件格式 (File format): CSV. (Note: unified reporting delivers CSV, whereas the old SP Advertised Product report was XLSX — see §6.6.)

F. Submit — 提交 button id=urc_frb_create_report_button (top-right; real click, not .click()). Cancel: id=urc_cancel_subscription_button. On success the builder closes and you return to /reporting with the new row in 正在处理 (Processing). If required columns are missing you get an inline banner "Fix the following issues and submit again" listing the missing time/detail/metric.

6.5 Standardized metrics & terminology — READ BEFORE trusting numbers

Unified Reporting renamed and redefined metrics. The underlying measurement logic is unchanged, but display names — and some values — differ from the old reports. Key changes:

• Common terms unified: (DSP → unified) 订单→广告活动, 广告订单项→ 广告组, 创意→广告, 点击跳转量→点击量.
• Base conversion metrics now include halo. 购买量 / 销售额 / 已售 商品数量 (and 商品详情页浏览量 etc.) now count both promoted products and brand-halo (品牌光环) products. To reproduce the old SP "promoted-only" numbers, use the （推广） variants (推广商品的购买量 / 推广商品的销量 / 已售商品数量（推广） / 推广商品的 ROAS). DSP users especially: dropping the old "总" prefix means base metrics can return higher values than the legacy DSP report.
• Lookback window removed from names. Base metrics automatically use the correct window: Sponsored Products seller = 7-day, vendor = 14-day; other campaign types = 14-day (flexible-lookback DSP uses the advertiser's chosen click/view windows). So 销售额 = the old "7-day total sales" for an SP seller; ROAS = ROAS7d.
• Traffic-date reporting. All conversions are now reported on the traffic date (the ad-interaction date), not the conversion date (DSP/legacy SB used to report on conversion date). Expect higher conversions early in a campaign's run vs the old reports.
• Off-Amazon conversions all carry the （亚马逊站外） qualifier.
• Qualifiers you'll see: （推广） promoted, （光环）/（品牌光环） halo, （品牌新客） brand-new-customer, 归因于点击/归因于浏览 click-/view-attributed, （已合并） on+off-Amazon combined.

6.6 Reproducing the old SP "Advertised product" report (monthly export)

The monthly 下载上月数据报表 schedule needs a per-marketplace "Sponsored Products Advertised product" report. In Unified Reporting that is the 推广的商品 template (advertised_product) filtered to 广告产品 = 商品推广 (SP) and one marketplace. The old XLSX had 24 columns; here is the mapping to unified dimensions/metrics (SP seller, 7-day window):

Old column (XLSX)	Unified dimension / metric
开始日期 / 结束日期	dim 日期范围 (summary) — or 日期 for daily rows
广告组合名称	dim 广告组合名称
货币	dim 广告活动货币代码 (or infer from 国家/地区)
广告活动名称	dim 广告活动名称
广告组名称	dim 广告组名称
国家/地区	dim 国家/地区
广告SKU	dim 推广的商品 SKU
广告ASIN	dim 推广的商品 (编号 = ASIN)
展示量	metric 展示量
点击量	metric 点击量
点击率 (CTR)	metric 点击率
单次点击成本 (CPC)	metric CPC
花费	metric 成本
7天总销售额	metric 推广商品的销量 (promoted sales)
ACOS 总计	derive (成本 / 推广商品的销量) — or an ACOS metric
总 ROAS	metric 推广商品的 ROAS
7天总订单数(#)	metric 推广商品的购买量
7天总销售量(#)	metric 已售商品数量（推广）
7天的转化率	metric 购买率（推广的商品）
7天内广告SKU销售量(#)	metric 已售商品数量（推广）
7天内其他SKU销售量(#)	metric 已售商品数量（光环）
7天内广告SKU销售额	metric 推广商品的销量
7天内其他SKU销售额	metric 销售额（光环）

Column order and exact header text will differ from the legacy XLSX (Amazon standardized the names), and the file is CSV, not XLSX. Downstream profit analysis keys on SKU/ASIN + spend + sales + units — which all map above — but if a consumer parses by exact legacy header string or .xlsx extension, it must be updated for the unified CSV. The data (per SKU × campaign × country: spend, sales, units, impressions, clicks) is fully reproducible.

Verified live (one SP marketplace, last-month period): the advertised_product template seeds ~47 default columns — including 推广的商品编号 (=ASIN), 推广的商品 SKU, 广告活动名称, 广告组名称, 广告组合名称, 预算货币, 展示量, 点击量, 总成本, and the base / （推广） / （光环） / （品牌新客） conversion families — and returns one row per SKU × campaign. Two things to know:

• The default set uses 推广的商品站点 (Advertised product marketplace, values AMAZON_SA/AMAZON_AE/AMAZON_AU), not 国家/地区. The report is account-wide — it contains every marketplace, and the country filter does NOT scope it (see §6.8). Split per-marketplace downstream on 推广的商品站点 (or add the 国家/地区 dimension, §6.4C, and split on that).
• The 日期范围 dimension shows each row's active date span within the selected period (rows with activity across the whole month show <1st> - <last>; sparser SKUs show a sub-span). The report period itself is still the one you picked (e.g. 上个月 = the full previous calendar month).

Example placeholder row (never use real SKUs/ASINs/campaign names): WIDGET-001-White, B0EXAMPLE001, "example manual - KSA", SAR, 沙特阿拉伯, 5000 impressions, 30 clicks, 30.00 spend, 250.00 sales, 6 units.

6.7 Retrieving a finished report

After 提交 the report runs once automatically. Poll /reporting (reload and read 报告状态) until it flips 正在处理 → 已完成 (Completed). Do not sleep >270 s at a stretch (the store session recycles at ~8 min idle); reload on a ~60–90 s cadence. Reports typically finish in a few minutes to ~30 min depending on data volume; the migration guide warns data can lag and be revised within the attribution window.

When 已完成, get the download from the report detail page, NOT the list-row action menu. The row's 操作 button (id=report-menu-trigger) menu only has 复制 (Copy) / 删除 (Delete) / edit — no download. Instead click the report name link → detail page /reporting/history?entityId=<ENTITY>&reportId=<report-uuid>, which has a 下载 (Download) link (data-takt-id=storm-ui-link). Its href is:

/reporting/subscriptions/<report-uuid>/download-report/<run-uuid>

Trigger the real anchor (JS .click() on the <a>, or browser-use open its href; a bare label click fires no download). The CSV lands in the store's download dir (~/.vibe-seller/downloads/<slug>/); glob the newest *.csv (Amazon names it after the report — e.g. <report-name>.csv), then mv/rename per the task's convention. If you set an email recipient, the CSV also arrives by email link.

6.8 Per-marketplace: split downstream (the filter does NOT scope)

Each marketplace has its own campaigns and its own currency (sa=SAR, ae=AED, au=AUD, …). Unified Reporting is account-wide, not per-marketplace, and — unlike the old per-TLD /reports console — there is no reliable way to get a single-marketplace file from the builder:

⚠️ Verified live (twice, identical MD5): neither the 国家/地区 filter nor the marketplace entityId scopes the report. Entering via a marketplace's entityId (e.g. the AE entity) still returns all marketplaces, and the 筛选条件 → 国家/地区 storm-ui dialog (全部清除 → tick one country → 保存) does not commit to the backend — the CSV still contains every marketplace. A report created for "SA" and one created with an "AE-only" filter came out byte-identical.

So: produce ONE account-wide report and split it downstream by marketplace. Add/keep the 推广的商品站点 (Advertised product marketplace) column — its values are AMAZON_SA / AMAZON_AE / AMAZON_AU — (or add the 国家/地区 dimension) and partition the rows into each country's folder yourself:

import pandas as pd
df = pd.read_csv('unified_advertised_product.csv')
col = [c for c in df.columns if '站点' in c or 'marketplace' in c.lower()][0]
for mkt, sub in df.groupby(col):        # AMAZON_SA / AMAZON_AE / AMAZON_AU
    cc = mkt.split('_')[-1].lower()      # sa / ae / au
    sub.to_csv(f'reports_{{MM}}_{cc}_{{slug}}/Sponsored_Products_Advertised_product_report_{{Mon}}.csv', index=False)

Verify content, not just that a file exists: after splitting, confirm each per-country file's 推广的商品站点 (or 国家/地区) values are all the intended marketplace. Never hand-copy one marketplace's rows into another country's folder — the currency and campaigns differ.

6.9 Limits & troubleshooting

• Report > 25 GB times out. Shorten the date range, keep advertisers < 50, or select fewer dimensions.
• Empty fields / empty report are normal / success, not failure — an inapplicable dimension, an unsupported historical lookback, or a permission-masked metric all yield blank cells. A 0-row export for a valid request is a completed deliverable (see the applicability rule near the top of this skill).
• Account not visible? You need access to the manager/advertiser account; newly-linked accounts can take up to ~1 hour to appear.
• Not yet in unified reporting (planned late-2026): MRC viewability, benchmarks, conversion path, SP video, SP prompts, cross-retailer SP, and legacy-SB targeting/search-term support. For those, use §6.10 until cutoff.

6.10 Legacy Sponsored-ads console (retiring 2026-12-31)

Only for reports not yet in unified reporting, or stores mid-migration — and only until the cutoff. The old console is at advertising.amazon.{tld}/reports (note: /reports, not /reporting). Its create dialog used #advertising-reports button → category/type controls (button[id*=report-type-control], types searchTerms, keywords, adProducts, campaigns, placements, purchasedProducts, …), a input[name=time-units] Summary/Daily radio, and #urc_run_subscription_button to run; the list page linked to a detail page whose history table exposed the Download link. Full legacy selectors are in this file's git history (pre-unified-reporting revision) if a mid-migration store still needs them. Do not build new workflows on /reports — it stops accepting new reports on 2026-12-17.

---

7. Download Behavior

Where Files Go

Browser-use downloads go to the browser profile's download directory, which is per-store. The path depends on the browser backend configured for the store (check stores/{store-slug}/STORE.md for the download path).

Files do NOT download to the task workspace automatically.

Copying Downloads to Task Workspace

The browser download directory is SHARED across all tasks. It accumulates files from every run. When copying to the task workspace:

• Only copy files matching the invoices/reports you just downloaded
• Files with (1), (2) etc. suffixes are re-downloads from previous runs. Always copy the latest version (highest number or most recent timestamp) and rename to remove the suffix: INV-001 (3).pdf → INV-001.pdf
• Do NOT copy the entire directory or use wildcards like cp *.pdf

# List downloaded files (newest first)
# DLDIR path depends on the browser backend configured for the store
ls -lt "$DLDIR/"

# Copy specific file to task workspace
cp "$DLDIR/INV-001.pdf" workspace/invoices/

# If only "(N)" versions exist, copy the latest and rename:
cp "$DLDIR/INV-001 (2).pdf" workspace/invoices/INV-001.pdf

Download Mechanism by Report

Report	Format	Mechanism	API Endpoint
Business Reports	CSV	Button click → direct download	Internal
Fulfillment	CSV/TSV	Request → generate → download	Internal
Tax Document Library	PDF	"View" button → form POST	POST /tax/seller-fee-invoices
Custom Reports	CSV	Request → generate → download	Internal
Payments Repository	CSV	"Download CSV" button → XHR	GET /payments/reports/api/download-report?reportId={UUID}
Advertising Reports (Unified Reporting)	CSV	report row 已完成 → 下载 link (real <a> click) → direct	Internal (/reporting)

Download Troubleshooting

If downloads don't appear in the browser download directory:

1. Check that the CDP proxy is correctly forwarding download headers
2. Try fetching the download URL directly via js("...fetch()...")
3. For Payment reports, the API endpoint can be called directly from the browser context
4. For Tax PDFs, the form POST may open in a new tab instead of downloading

---

Quick Reference: Navigation Cheat Sheet

Seller Central Home
├── hamburger menu (☰) → hover "Reports"
│   ├── Business Reports        → /business-reports
│   ├── Custom Analytics        → (new)
│   ├── Fulfillment             → /reportcentral/WelcomePage
│   ├── Advertising Reports     → advertising.amazon.{tld}/reporting (Unified; old /reports retiring 2026-12-31)
│   ├── Return Reports          → /reportcentral/WelcomePage
│   ├── Custom Reports          → /listing/reports
│   ├── Inventory Reports       → /inventory-reports
│   ├── Tax Document Library    → /tax/seller-fee-invoices
│   ├── Manage Taxes            → /tax/tax-settings
│   └── Selling Economics & Fees → /selling-economics-and-fees
│
├── hamburger menu (☰) → hover "Payments"
│   └── Payments Dashboard      → /payments/reports-repository (tab)
│
└── Advertising Console (separate app)
    └── 效果衡量和报告 (Measurement & reporting) → 报告 (Unified Reporting)

Tips

• Always dispatch mouseover on hamburger menu categories, never a click
• Check aria-expanded in page_info() output to confirm the submenu is revealed
• For instant reports (Business, Tax Library): click Download/View directly
• For generated reports (Fulfillment, Custom, Payments): Request → poll status → Download when ready. Typically 1-5 minutes.
• For ad reports (Unified Reporting, §6): 使用模板/创建报告 → 提交 → poll 报告状态 until 已完成 (few min – 30+ min) → 下载
• Always copy downloads from the browser download dir to task workspace
• Use direct URLs when possible to skip hamburger menu navigation
• The hamburger menu button is inside a navigation-hamburger-menu shadow DOM element — look for role=button inside it
• Payment CSV: skip first 7 lines (definitions) when parsing
• Ad report CSV: no header lines to skip, starts with column row
• When building analysis scripts, use the column structures documented above