advertising-ops

Advertising Ops is a Claude Code skill that automates the process of finding proven competitor ads, tearing them down, and generating new ad creative based on what is already working. It is aimed at founders, agency operators, and media buyers who want to model their ads on what the market has already validated rather than starting from nothing.

The tool runs as a single command inside Claude Code and works in five phases. It starts by gathering context about your business, either reading an existing brand kit file or asking you questions about your product, target customer, and keywords. It then searches the Meta Ad Library, the public database of ads running on Facebook and Instagram, filtering for ads that started at least two months ago and are still running today. The reasoning is that an advertiser still spending money on an ad after months has received enough return to keep it going, which makes longevity a proxy for proof.

For video ads, the tool downloads each video, extracts frames using ffmpeg, transcribes the voiceover, and analyzes the structure: specifically the opening hook, the narrative flow, and the call to action. The README notes that most of an ad's effectiveness lives in the first few seconds and the script, not just the visuals.

After the research and teardown phases, the tool runs a structured conversation to pin down your exact offer and a single call to action before generating anything. The final output is at least five ad variations, each in its own folder containing a generated image or video and a text file with the aligned copy.

Three external services are required at runtime: Apify for scraping the Meta Ad Library, Higgsfield for image and video generation, and ffmpeg for video frame extraction. Installation is a single npx command.