How to use the Apple Ads Audit agent skill (ads-apple)

What is an agent skill and what does ads-apple make the agent do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent reads when a task matches the skill's description. The ads-apple skill tells the agent how to review an Apple Ads account against a fixed set of areas: AdServices and AdAttributionKit measurement, attribution reconciliation, campaign and keyword structure, Search Match, App Store placements, custom product pages, bids, budgets, and policy.

Before reviewing, the agent collects context such as the objective, the conversion definition, account and campaign age, geography, date window, timezone, currency, spend, targets, and which data sources are available. It then returns observations, diagnoses, prioritized recommendations, unscored opportunities, contradictions, and missing inputs. Any proposed account change stays a draft until it passes a separate approval gate; the skill does not apply changes to live campaigns.

Install the ads-apple skill

Download the ZIP for skills/ads-apple from the item page and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. Either way, the skill's SKILL.md should sit inside its own folder in the skills directory.

If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

Check the requirements before you start

The skill refers to a parent ads/ plugin directory that holds the main operating contract and reference files, including ads/references/apple-audit.md. That directory is not included in the single SKILL.md file, so a standalone install may not have everything the procedure asks the agent to read. The skill says to resolve these files from the plugin root or from the skills directory in a standalone install.

You also need to supply the data. The skill works from account exports, screenshots, or API results from Apple Ads that you provide. The results depend on that data, and the skill notes that it may be incomplete or inaccurate.

• Apple Ads exports, screenshots, or API results covering the period you want reviewed
• Your conversion definition, objective, geography, date window, timezone, and currency
• Access to the ads/ reference files if you want the full checklist

Trigger the skill with natural requests

The skill activates when your request matches its description, which mentions Apple Ads, Apple Search Ads, App Store ads, Search Match, custom product pages, AdServices, or Apple app-install campaigns. You do not need a special command. Say what you want reviewed and attach the data.

Good requests state the goal and the scope. For example, you might ask the agent to audit the attribution setup for an app-install campaign, or to check whether Search Match and keyword structure are working against your stated goal. Include the date range and the conversion event you count as a success, because the skill asks for these before it evaluates anything.

• Audit my Apple Ads account for an app-install campaign. Here is the export for the last 30 days and my conversion definition.
• Check my AdServices and AdAttributionKit setup and reconcile it with my MMP numbers.
• Review my custom product pages and App Store placements for this campaign.
• Look at my bids and budgets and tell me what to test first, as drafts only.

Read the output with care

The output separates what the agent observed in your data from its diagnoses and from its recommendations. It also lists contradictions and missing inputs, which are often the most useful part because they show where your data cannot support a conclusion. Features the skill treats as optional, beta, premium, unavailable, or ineligible are kept unscored.

The skill warns against applying benchmarks without checking the objective, geography, methodology, sample size, conversion lag, and account maturity. Treat any number it cites as a reason to check your own data, not as a universal target.

• Observations: what the data shows
• Diagnoses: likely causes, with uncertainty marked
• Recommendations: prioritized, kept as drafts
• Contradictions and missing inputs: what to fix in your data first

Tips for a useful Apple Ads audit

Export the data for the same date window and timezone you use for reporting, and keep the currency consistent. Note the conversion lag so recent days are not judged too early. If you reconcile with a mobile measurement partner, provide both the Apple numbers and the MMP numbers so the agent can flag differences.

Run the audit again after changes to see whether the same findings persist. Because the skill returns drafts, review each proposed change in Apple Ads yourself before making it.

• Use one timezone and currency across all exports
• Share MMP and Apple numbers together for reconciliation
• Keep the review scope narrow (for example, one campaign group) if the data is large
• Verify every recommendation against your own goals before acting

Safety notes: scripts and network access

The skill does not run scripts and does not use the network, according to its listing. It works from the files and data you give it. Because external account content is treated as data rather than instructions, pasted text from reports or web pages should not be able to direct the agent. You still remain responsible for every change made in your Apple Ads account.

• Runs scripts: no
• Uses network: no
• Proposed changes are drafts until a separate approval gate passes