meta-ads-creative

Files (2)

• SKILL.md
• evals.json · evals/

SKILL.md (first lines)

# Meta Ads Creative Studio

Create a reviewable creative slate from real customer language and account evidence. This skill produces briefs and copy; it does not imply that the NotFair Meta MCP can upload creatives or create campaigns.

## Setup

Read and follow `../shared/preamble.md`. Read `{data_dir}/meta/business-context.json` and `{data_dir}/meta/personas/{accountId}.json` first. If either is missing or stale, recommend `/meta-ads-audit` before writing. The audit's creative inventory and personas are the baseline; do not invent an audience because a request is underspecified.

Pull current ads and ad-level insights using available read capabilities when account access exists. Use link CTR, CPA or purchase value, frequency, attribution window, and a named time period. If the user has reviews, sales-call notes, or customer research, treat them as evidence only when the source is identifiable.

## Create a concept slate

Each concept must be a real hypothesis:

`persona × motivation × angle × format × one variable to test`

Start with three distinct concepts, not three cosmetic versions of the same idea. A useful spread is proof, problem/pain, product demonstration, founder/expert explanation, or an approved offer — but select only angles supported by the evidence.

For every concept, include:

| Field | Required content |
|---|---|
| Concept ID and hypothesis | Who it is for, why it should work, and the single variable under test |
| Evidence | Named campaign/ad/customer source, metric and window, or approved business-context field |
| Hook and primary text | First-frame or first-line hook plus placement-adapted copy |
| Visual brief | Subject, setting, action, proof shown on screen, and production notes |
| CTA and destination | CTA plus approved landing page/message-match note |
| Claim ledger | Source and approval status for every rating, result, price, guarantee, quote, or before/after claim |
| Read plan | Primary metric, guardrail, planned duration/exposure, and what would falsify the hypothesis |

An unsupported claim is `needs_substantiation`, not ad copy. Supply a truthful alternative rather than making a vague disclaimer. Do not fabricate UGC testimonials, review counts, customer outcomes, or visual proof.

## Creative refresh and testing

For fatigue, first verify the signal: compare link CTR and CPM week over week, cite frequency, and preserve the attribution window. Frequency rising with a material decline in link CTR is a refresh hypothesis, not a universal rule.

Test one strategic variable at a time: angle, hook, proof, format, audience, or destination. Set the winner metric and guardrail before launch; do not name a winner from an early CTR spike or an unqualified in-platform ROAS number. Keep a short iteration log: concept ID, launch date, audience, spend, results, decision, and next challenger.

## Production and execution boundary

Deliver a production-ready brief for the user's approved workflow. If publication is requested, check whether the live connection supports the specific creation or upload operation and follow its current contract. Before a human publishes, require review of claim evidence, rights/releases, destination, policy-sensitive targeting, and placement-safe crops.

Use `/meta-ads` to diagnose performance or act within its supported mutation surface. Use `/meta-ads-audit` when context, tracking, or creative inventory is missing.

From Swastik023/marketing-api-automation-mcp, MIT.