finds.dev← search

// the find

AgriciDaniel/claude-ads

★ 5,985 · Python · MIT · updated Jun 2026

Comprehensive paid advertising audit & optimization skill for Claude Code. 250+ checks across Google, Meta, YouTube, LinkedIn, TikTok, Microsoft & Apple Ads with weighted scoring, parallel agents, industry templates, and AI creative generation.

A Claude Code skill that runs structured paid advertising audits across 8+ platforms (Google, Meta, LinkedIn, TikTok, etc.), outputting a 0-100 health score and prioritized action plan. It's a prompt-and-reference-file package dressed up as a tool — there's no code doing the analysis, Claude is doing it guided by SKILL.md files. Target audience is PPC agencies and consultants who want a repeatable audit framework without a $500/mo SaaS subscription.

The check catalog is bidirectionally verified against test fixtures in CI — 209 checks can't drift from the reference files without breaking a test, which is unusually disciplined for this category of repo. Platform coverage is current: Andromeda creative-similarity detection, AI Max negative-keyword gaps, and AdAttributionKit configurable windows are 2025-2026 changes that most audit checklists haven't caught up to. The install scripts include SSRF regression tests and whitelist validation on target paths — someone thought about security rather than just piping curl into bash. The public/private mirror split is a honest funding model: MIT public release, paid community gets early access.

The '250+ checks' claim is marketing math — only 209 are actually verified in CI; the rest are inline thresholds in SKILL.md files with no test coverage. The core product is LLM-guided checklist execution, which means output quality depends entirely on what data you paste in; without live API access the 'audit' is really 'Claude reading your CSV export.' The optional MCP integrations for live data are third-party tools with their own auth complexity, and the README's self-hosted Meta option is hand-waved as 'wrap the Marketing API yourself.' Single maintainer with no SLA on features is a real adoption risk for agencies building workflows on top of this.

View on GitHub → Homepage ↗

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →