# Create Grounded Ad Creative at Scale

Ad teams need many platform-ready variations without losing evidence, brand fit, or review clarity. This skill creates grounded copy, concepts, test plans, and review pages.

## Install

```bash
npx skillstore add coreyhaines31/ad-creative
```

## Metadata

- Status: approved
- Slug: coreyhaines31-ad-creative
- Version: 2.7.0
- Author version: 2.7.0
- Skillstore revision: r1
- Version status: valid
- Tree hash: 43c48b34179d2ef96c258ca8e7613f10d981ec6593b100d61795887aee66cf07
- Author: coreyhaines31
- GitHub username: coreyhaines31
- License: MIT
- Repository: https://github.com/coreyhaines31/marketingskills/tree/main/skills/ad-creative/
- Ref: 635f69fb8d2f4e6330ba47a4e5a0fb239c04d110
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: scripts, external\_commands, filesystem, network, env\_access
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/coreyhaines31-ad-creative
- Manifest: https://skillstore.pages.dev/api/skills/coreyhaines31-ad-creative/manifest

## Capabilities

- Generates headlines, descriptions, primary text, and complete variations for major paid advertising platforms.
- Checks copy against documented platform character limits and format requirements.
- Analyzes provided performance data to identify winning themes and create controlled iterations.
- Builds grounded static concept batches from winning ads, reviews, comments, and brand assets.
- Creates evidence-ranked creative roadmaps, testing plans, iteration logs, and retrospective summaries.
- Populates a self-contained HTML review page for stakeholder comparison and approval.

## Use Cases

- Launch a Paid Campaign: Create platform-compliant ad variations from a product brief, audience definition, offer, and brand voice.
- Iterate from Performance Data: Find patterns in supplied campaign results and generate focused variants that preserve winning angles.
- Review a Creative Batch: Turn grounded concepts into a visual review page that clients can compare before production or launch.

## Prompt Templates

### Create a Starter Ad Set

```
Create five Meta ad concepts for [product] aimed at [audience]. Use [offer] and this brand voice: [voice]. Ask for missing essentials first.
```

### Build Platform Variations

```
Use this product context: [context]. Generate Google, Meta, and LinkedIn variations. Show each angle, character count, CTA, and required disclaimer.
```

### Iterate from Results

```
Analyze this ad performance table: [data]. Identify strong and weak patterns. Create twelve controlled variations and state the hypothesis behind each group.
```

### Plan a Grounded Creative Cycle

```
Use these winning ads, reviews, comments, and account results: [sources]. Build an evidence-ranked roadmap, production mix, review page plan, and monthly retro framework.
```

## Limitations

- It cannot retrieve campaign performance without configured external tools, credentials, and user approval.
- Output quality depends on accurate product context, customer evidence, brand guidance, and performance data.
- Generated ads do not guarantee performance, legal compliance, or platform approval.
- Image, video, and voice production may use external services requiring consent, licensing, privacy review, and separate fees.

## Best Practices

- Provide platform, audience, offer, product proof, brand voice, and compliance constraints before generation.
- Ground claims in current reviews, comments, winning ads, and verified product facts.
- Change one major variable per test and preserve a clear hypothesis for every variation group.

## Anti Patterns

- Do not invent statistics, testimonials, endorsements, urgency, or product capabilities.
- Do not generate many near-duplicate lines that test wording instead of distinct strategic angles.
- Do not upload or publish creative without human review for claims, rights, consent, and platform policy.

## Security Audit

- Audited at: 2026-07-13T17:46:52.014\+00:00
- Summary: All 123 static alerts are false positives caused by escaped HTML rendering, JavaScript template literals, Markdown, relative links, and documented provider API examples. One high-severity concern remains: voice-cloning guidance does not require consent or authorization, which can enable impersonation. No prompt injection or credential exfiltration evidence was found.

## Stats

- Views: 112
- Downloads: 9
- Favorites: 0
- Popularity score: 0