agentleFS
slilbudgetslilbudget/knowledge-delta-skills/skills/google/google-feed-ops/SKILL.md

google-feed-ops

Skill3 starsChanged 49 days agoMITView on GitHub

Google Merchant Center (GMC) launch and operation plus the retail data layer: new-account launch sequence (site trust → ToS → verify+claim → business info → shipping/tax → data source → review → programs → Ads link → first campaign), Merchant API v1 mechanics and the gmcops CLI, feed spec and attributes, feed rules and supplemental sources, titles, custom_label schema, GTIN, suspensions and appeals, MC↔Ads link, free listings, CSS. Use for: 'set up Merchant Center for a new store', 'products not showing / disapproved', 'GMC suspended Misrepresentation', 'push products via API', 'link Merchant Center to Google Ads'. Campaign types and PMax bidding live in google-ads; grey cascade in google-grey-ops.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Content API for Shopping sunset 2026-08-18: Only API integrations broke — manual upload, scheduled fetch, Sheets, and platform connectors …

    line 42
  2. Very specific:

    Never fake a gtin or mpn with an internal SKU — set identifier_exists=no.

    line 80
  3. Very specific:

    Never submit a sale_price without the real non-sale price.

    line 81
  4. Distinctive:

    Use the test/preview mode on attribute rules.

    line 83

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Google Feed Ops

Reviewed 2026-09-02. Baseline: Sonnet 5 / Claude Code subagent / 2026-09-02; 2026-09-03 blind suspension task [R] — site-review and second-account rules already known. The retail data layer. Google Ads buys the traffic; the feed decides what is eligible, what it costs, and what gets suspended — and the Merchant Center account is reviewed on the site, not the feed. No Facebook analogue at this depth.

Route: buy mechanics → google-ads · infra/survival/agency accounts and MC↔Ads cascade → google-grey-ops (12) · counting → tracker-ops · portfolio decisions → senior-buyer-ops.

Check first

NeedReference
New account → products serving, in order; first campaign on a fresh pairreferences/04-gmc-launch-runbook.md
Merchant API v1 facts, what the API cannot do, gmcops doctor gatesreferences/05-merchant-api-ops.md
Attributes, limits, submission methods, titles, custom labelsreferences/01-feed-spec-and-submission.md
Suspensions, the gates that block review, required website elements, ratingsreferences/02-merchant-center-suspensions.md
Standard Shopping priority ladder, PMax retail structure, product-level reporting, CSSreferences/03-shopping-pmax-retail.md

gmcops lives in google-grey-ops/scripts/ (one uv project with googleops); usage in google-grey-ops/10. Read-only doctor first: gmcops --account <id> --json doctor --country US.

Three things that break accounts

  1. The five review gates. Shipping settings · a data source for the target country · website URL claimed and verified · business address verified · US tax settings. Missing any one blocks review entirely. This is the real cause of most "products not showing" tickets — check before diagnosing anything else.
  2. Feed vs live page mismatch on price and availability. Google's automated crawl fires preemptive item disapproval. It is automated, so the fix is a sync fix, never an appeal-and-argue. Compare against the checkout total, not just the product page.
  3. Promotional text in titles or descriptions. A policy trigger, not a soft ranking issue.

Migration status 🔺

  • Content API for Shopping sunset 2026-08-18. Only API integrations broke — manual upload, scheduled fetch, Sheets, and platform connectors were unaffected.
  • Merchant API v1beta discontinued 2026-02-28, earlier than the Content API sunset.
  • Current: merchantapi.googleapis.com, stable v1.
  • UCP (in-Google checkout via native_commerce(checkout_eligibility)) is early access, select merchants, US/CA/AU only. It did not replace Buy on Google, which shut down 2023-09-26.

Model

Merchant Center account -> verification gates -> data source(s)
  primary feed (structured) -> feed rules / supplemental feeds (overlay)
    product -> id / title / price / availability / GTIN / custom_label_0-4
      listing group filter -> asset group (PMax) or product group (Shopping)
        -> bid strategy per segment

The throughline: label schema → listing-group filter → divergent bid strategy per filter — how per-segment economics happen without per-SKU bidding.

Workflow

  1. Clear the five review gates before anything else.
  2. Establish one primary submission method per data source — mixing causes silent last-write-wins conflicts.
  3. Fix required attributes and GTIN coverage — a missing real GTIN costs visibility and disables price benchmarking.
  4. Build the custom_label schema before campaign structure, since structure depends on it: margin tier, price band, bestseller rank, seasonality, stock lifecycle (canonical table in 01, don't reinvent).
  5. Optimize titles for the ~70-character cliff, specs before adjectives, decision-driving attributes first.
  6. Confirm destinations and listing-group keys are valid. PMax/Shopping campaign structure, feed-only vs assets, brand exclusions, and product-level reporting live in google-ads (01, 07).

Guardrails

  • id is the join key. Never reuse or recycle it.
  • Never fake a gtin or mpn with an internal SKU — set identifier_exists=no.
  • Never submit a sale_price without the real non-sale price. That is a misrepresentation trigger.
  • shipping_label and custom_label_0-4 are unrelated. Conflating them is a common integration bug.
  • Use the test/preview mode on attribute rules. It prevents catalog-wide breakage.
  • Campaign priority does nothing unless two campaigns share the same products in the same country and language — and an under-budgeted high-priority campaign silently falls back to the low-priority one.
  • AI title generation belongs in a testing sandbox, not in production. Codify winning patterns into templates with fallback chains.
  • CSS is EEA/Switzerland/UK only. The ~20% figure has zero US applicability — never cite it to a US advertiser.
  • Merchant Center and Google Ads suspensions are separate systems with separate appeals, but a linked pair is one blast radius: a suspended Ads account linked to the MC suspends the MC, and unlinking afterwards does not clear it (official). Fix the Ads side first, then request MC review.
  • Never open a second MC account to escape a suspension — same domain/store/identity gets suspended again and reframed as Circumventing systems.
  • Fix everything before requesting review. 1–3 attempts before the button locks; each denial adds a growing cool-down. issueresolution.triggeraction (programmatic appeal) is allowlist-gated.
  • Products are only writable via API into API-type data sources; processed products appear after several minutes; the key is contentLanguage~feedLabel~offerId.
  • US tax settings, phone/address verification, and identity documents have no API surface — UI.

Output

Lead with which layer the problem sits in — gates, feed data, policy, or campaign structure. Most "campaign" problems in retail are feed problems. Give the exact attribute or setting, the expected review timeline (3–5 business days initial, up to 7 for account suspension, 24–48h for a re-crawl), and what to verify after the next refresh.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.