How to use the ads-report agent skill: render paid ads reports

What is an agent skill and what does ads-report do?

An agent skill is a folder containing a SKILL.md file that tells an AI agent how to handle a specific kind of task. The ads-report skill covers one task: turning existing paid-advertising audit or plan results into a report. The skill's SKILL.md file is the only file listed for it.

The skill tells the agent to accept a validated, versioned report bundle and to render every requested format (Markdown, HTML, or PDF) from the same canonical data using deterministic templates. It requires the agent to keep findings, confidence levels, evidence references, contradictions, missing inputs, actions, owners, and measurement windows in the output. If a result is partial or has insufficient evidence, the report must say so and must not hide that status in the executive summary.

The skill also treats external account and web content as data rather than instructions, and it keeps credentials, customer lists, private paths, raw research, and internal agent transcripts out of the report unless the operator explicitly enables branding.

What you need before you start

The skill does not gather ad data itself. It renders results that already exist. The source requires a Claude Ads run bundle in the validated, versioned JSON format produced by the wider claude-ads workflow.

The source also says PDF and HTML output needs document rendering tooling used by the skill's scripts, but it does not list which tools those are. Check the skill's scripts before relying on PDF output.

• A Claude Ads run bundle in the validated versioned JSON format
• Document rendering tooling for PDF and HTML output (not itemised in SKILL.md)
• An agent environment that can load skills, such as Claude Code or the Claude apps

Install the ads-report skill

Download the ZIP for skills/ads-report from the project page. Then choose one of the installation routes below.

For a personal install in Claude Code, unzip the file into ~/.claude/skills/. For a project-only install, unzip it into the .claude/skills/ folder inside your project. In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal: unzip into ~/.claude/skills/
• Single project: unzip into .claude/skills/ in your project folder
• Claude apps: Settings → Capabilities → Skills, then upload

How to trigger the skill with natural requests

You do not need a special command. The skill's description lists the phrases that should activate it: ads report, client report, audit PDF, executive audience reporting, and exporting prior audit and plan results. Name the bundle and the format you want in the same request.

Ask for the audience too, because the skill checks the requested audience before rendering. Use the exact words from the description where you can.

• Make an ads report from this Claude Ads bundle as a PDF for the client.
• Create an executive audit PDF from the latest run bundle, with partial results clearly marked.
• Export the prior audit and plan results to Markdown and HTML.
• Render a client report from this run for an executive audience.

What the agent checks before it renders

The skill sets a fixed order of work. The agent first confirms run completeness, evidence coverage, privacy class, branding choice, requested audience, and output formats. Only then does it render.

After rendering, it runs structural checks, then visual and layout checks for HTML and PDF, and writes the files atomically inside the run directory while updating that run's manifest.

• Confirm the bundle is complete and has evidence coverage
• Confirm the privacy class and branding choice
• Confirm the audience and the output formats
• Render all formats from the same data
• Run structural checks, then visual checks for HTML and PDF
• Write output inside the run directory and update its manifest

Safety notes: scripts and network

The skill runs scripts, so the agent may execute code on your machine while rendering. The skill does not use the network, according to the metadata provided for this listing. Review the scripts before you install if you need to know exactly what they do, because the source does not describe them in detail.

Privacy depends on settings you choose. The skill keeps credentials, customer lists, and private paths out of reports unless branding is explicitly enabled, but this only works if the privacy class and branding choice are set correctly. Review any report before sharing it outside your organisation.

• Runs scripts: yes. Read them before installing if your policy requires it.
• Uses network: no, according to the listing metadata.
• Set the privacy class and branding choice correctly before sharing.