Product Feed Optimizer
Audit and repair Shopping and Performance Max product feeds for approvals, accurate listings, stronger titles, and feed-based campaign structure.
The skill treats feeds, diagnostics, and scraped pages as untrusted input, forbids inventing identifiers, prices, stock, or specifications, and requires user confirmation before saving results. However, this skill does not fully specify permission boundaries, rollback, landing-page data flow, or confirmation controls for external pushes, so 7 points are deducted.
Inputs, outputs, workflow steps, NEEDS_INPUT handling, exception flags, and scope boundaries are largely consistent, with per-item reporting of missing or unresolved data. The static files provide no skill-specific key-path tests, real failure feedback, or execution reproduction, so the score remains below the static ceiling at 9.
Triggers, scenarios, input formats, expected output package, and boundaries from ad-copy and account-audit skills are reasonably clear. Chinese labeling exists, but the core workflow assumes Shopping, PMax, and Merchant Center ecosystems without evidence of mainland-China reachability or alternatives; platform-specific and non-fit boundaries are also limited, so 6 points are deducted.
The documentation is well layered and includes Quick Start examples, a skill contract, procedures, quality gates, references, version, Apache-2.0 licensing, and next-skill routing. Maintenance ownership, update commitments, changelog details, and some relative-path dependencies are not clear in the supplied material; publisher identity is also unverified, so 5 points are deducted.
The promised remediation package is concrete: per-item disapproval triage, title and description rewrites, attribute completeness, price and availability checks, and feed-based grouping. Because character limits and rules are marked Estimated and no real output or independent execution evidence is supplied, direct cross-platform effectiveness is unverified; the static score is capped at 6.
The skill supplies internal references, explicit rules, and one example marked Estimated and illustrative. Repository tests cover architecture and audit-artifact validation, not this skill's key paths or third-party platform facts. There is no independent reproduction or corroborating evidence, so 3 points are awarded.
- This is a static review: the skill was not executed, and platform rules or real feed outputs were not independently verified.
- The reference character limits and example are explicitly marked Estimated; validate them against current target-platform specifications and each actual landing page.
- Optional feed pushes or other external connectors should use separate confirmation, least privilege, and recovery controls; rollback is not fully defined by this skill.
- The core scenario relies mainly on Google/Meta-style overseas advertising ecosystems, with no evidence of mainland-China network reachability.
What does this skill do, and when should you use it?
Product Feed Optimizer is the paid-ads research skill in the Aaron Marketing Skills repository, focused on the product data behind Shopping and Performance Max campaigns. It reads product-feed exports, Merchant Center diagnostics, landing-page price and availability data, campaign goals, and target platforms. It produces disapproval triage, title and description rewrites, attribute-completeness checks, identifier and price/availability hygiene fixes, and feed-driven listing or asset-group structures. It fits teams repairing the foundation of product advertising, but it does not write text ads, score an entire account, or optimize landing pages.
Reads TSV, CSV, or XML product feeds; Merchant Center diagnostics or disapproval lists; catalog data; destination landing pages; campaign goals; and platform details. It identifies missing required or recommended attributes, GTIN or MPN issues, brand and category gaps, price or availability mismatches, image concerns, policy risks, and disapproval causes. It rewrites titles and descriptions using front-loaded high-intent attributes, then produces a disapproval triage table, attribute-completeness map, identifier/availability/price hygiene checklist, and listing-group or asset-group structure keyed to real fields such as product_type, brand, or custom labels. Missing evidence is marked as “[needs source]”; the skill does not invent GTINs, prices, stock, or product specifications.
- An ecommerce team is preparing Shopping campaigns and needs to validate required fields and price/stock consistency before launch.
- A paid-media operator has a Merchant Center disapproval export and needs item-level causes and proposed fixes.
- A brand wants product titles and descriptions rewritten so brand, product type, key specifications, and variants appear earlier.
- A Performance Max team needs listing-group or asset-group structures based on real product categories, brands, or custom labels.
- A growth team has detected differences between feed data and landing-page price or availability and needs a pre-launch reconciliation.
What are this skill's strengths and limitations?
- Covers disapproval triage, attribute completeness, title and description work, identifiers, price/availability, and campaign grouping in one workflow.
- Requires evidence from feed and landing-page data and explicitly forbids fabricated identifiers, prices, stock, or specifications.
- Supports Tier 1, keyless analysis using user-provided feed and diagnostics exports.
- Hands results into the repository's ROAS framework, with emphasis on the Offer dimension.
- Does not write RSA or other text-ad assets, score the full account, or compute RQS.
- Cannot fill factual gaps when feed, diagnostics, or landing-page evidence is missing.
- The source provides no standalone platform test matrix or dedicated test-suite evidence for this skill.
- Platform APIs are optional Tier 2/3 conveniences for pushing changes, not prerequisites for producing the remediation package.
How do you install this skill?
Clone the repository with git clone https://github.com/aaron-he-zhu/aaron-marketing-skills, then use ad/research/product-feed-optimizer/SKILL.md. For Claude Code, the repository documents /plugin marketplace add aaron-he-zhu/aaron-marketing-skills followed by /plugin install aaron-marketing@aaron. Compatible Agent Skills hosts can install the collection with npx skills add aaron-he-zhu/aaron-marketing-skills. The source does not document a separate standalone installation command for this skill.
How do you use this skill?
Provide a product-feed export and Merchant Center diagnostics, then trigger it with a prompt such as Audit my Shopping feed export for disapprovals and missing attributes: [path]. Goal is DR. or Rewrite these product titles to a front-loaded pattern and fill the missing GTIN/brand/condition attributes. [feed CSV]. If the feed or diagnostics are unavailable, supply those inputs first; also provide target platforms, the campaign goal, and landing pages for price and availability verification.
How does this skill compare with similar options?
Unlike ad-creative-builder, this skill works on product-feed data rather than text-ad assets. Unlike ad-account-auditor, it prepares and hardens the feed but does not score the account or run ROAS veto decisions. Unlike landing-optimizer, it verifies landing-page price and availability truth but does not modify the post-click page.