← ClaudeAtlas

ai-marketing-governancelisted

Use when a marketing team has adopted AI faster than anyone wrote rules for it — deciding which tools are approved, what must never go into a prompt, who reviews before publish, and how AI-drafted claims get fact-checked. Also use when the user mentions AI policy, AI governance, AI usage rules, AI disclosure, hallucinated citations, fake statistics, made-up sources, prompt data leakage, "is it safe to put this into ChatGPT", or reviewing AI-written marketing copy. Produces an adoption-readiness verdict, a completed policy with a named owner, and a per-claim citation check applied before anything publishes.
sarojkjha/aaj-marketing-skills · ★ 1 · AI & Automation · score 70
Install: claude install-skill sarojkjha/aaj-marketing-skills
# AI Marketing Governance AI can get a marketing team to a draft far faster than it could get there alone. What it cannot do is take responsibility for what goes out. This skill is the set of checks that sit between the draft and the publish button. One principle carries everything: **AI drafts. People decide and sign.** ## When to use The team has started using AI on company work with no written rules, a founder wants a policy they can adopt this week, someone is about to publish AI-drafted content, or a figure in a draft needs checking before it ships. Do not use this to write an aspirational AI ethics statement. A policy nobody owns is a document, not a policy — every section here ends in a name. ## Before you start 1. **Read the brand/product context first** (`.agents/product-marketing.md` / `.agents/aaj-brand.md`). 2. **Get the actual tool list,** by task and by person. "AI" is not a tool, and a policy that does not name them cannot be followed. 3. **Get the plan tier for each tool,** because whether it trains on inputs depends on it. Business and enterprise tiers usually do not; consumer tiers often do, and the default is rarely the safe one. If the user does not know, that is the first finding, not a blocker to note later. 4. **Get the reviewer's name.** Not "the team" — a person. ## Method Two failures are being prevented, and they are not the same. **Leakage** is what goes in. Treat every prompt as though it may be read by someone outside the compa