claude-ads

Claude Ads is a Claude Code skill that brings structured, evidence-backed paid-media operations to agencies, consultants, and in-house performance teams. It covers twelve advertising platforms—Google Ads, Meta Ads, YouTube Ads, LinkedIn Ads, TikTok Ads, Microsoft Advertising, Reddit Ads, Snapchat Ads, X Ads, Apple Ads, Amazon Ads, and Pinterest Ads—each with a focused skill, audit worker, control reference, capability declaration, and testable routing surface. The tool is read-only by default. Live account changes remain disabled until the exact platform and operation pass approval, idempotency, verification, audit, and rollback gates. This safety-first design means the tool can be used to analyze and plan without risk of unintended mutations. Core capabilities include: auditing paid-media accounts with dated evidence and explicit confidence levels; planning campaigns, channels, budgets, measurement, and experiments; creating copy, image, video, and product-photo briefs and assets; monitoring pacing, delivery, tracking, fatigue, policy, and performance; and producing versioned JSON reports that render to Markdown, HTML, and optional PDF. The architecture uses a single conductor that owns scope, policy, aggregation, and final artifacts, dispatching platform and cross-platform workers that analyze bounded slices and return schema-valid findings. Required-worker failure makes the run partial rather than silently presenting an incomplete audit as complete. Scoring separates health, evidence coverage, regulatory exposure, and opportunities. Controls use pass, fail, unknown, or not_applicable states. Coverage of 80% or more is graded, 60 to 79% is provisional, and below 60% is insufficient evidence. Optional, beta, premium, unavailable, and ineligible features stay unscored. A disabled or unapproved platform profile produces no health score. Account safety requires six conditions before any write: a tested and enabled capability for the exact operation, explicit account and object IDs, a human-readable before-and-after diff with blast radius, owner approval within account-defined ceilings, an idempotency key with audit destination and rollback, and verification that remote state still matches the mutation precondition. Missing ceilings mean no write. Permanent deletion is not supported in v2. The control-plane directory records product boundaries, dated sources, claims, capabilities, safety rules, privacy rules, ecosystem decisions, and release requirements. No source means no current platform claim. No implementation, fixture, and test means no capability claim. No approval and rollback means no account mutation. No independent verification means no release. Installation supports Claude Code as the canonical runtime, with Codex, Gemini, Cursor, Windsurf, Goose, and compatible Agent Skills hosts able to consume the same skill files. Managed dependencies support CPython 3.11 and 3.12 on Linux, macOS, and Windows. Browser capture requires an operator-installed Playwright browser payload, and PDF rendering requires WeasyPrint and Pango system libraries. The project is released under the MIT License, with a public release at Agricidaniel/claude-ads and a community mirror at AI-Marketing-Hub/claude-ads for early access and direct collaboration.