Ad Copy Generator
Generate multiple paid ad variants for Meta, Google, and LinkedIn with different hooks.
Installation
- Make sure Claude is on your device and in your terminal.
Skills load from
~/.claude/skills/when Claude Code starts up β so you need it on your machine first. If you don't have it yet, install it once with the command below, then runclaudein any terminal to verify.One-time setupnpm i -g @anthropic-ai/claude-codeAlready have it? Skip ahead.
- Paste into Claude Code or into your terminal.
This copies the whole skill folder into
~/.claude/skills/ad-copy-generator-marcosmodly/β the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Safe default: works for every skill.Faster alternative (instruction-only skills)
Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates β they won't be downloaded and the skill will fail when it tries to load them.
Quick install (SKILL.md only)Sign up to copy - Restart Claude Code.
Quit and reopen Claude Code (or any other agent that loads from
~/.claude/skills/). New skills are picked up on startup. - Just ask Claude.
Skills auto-activate when your request matches the skill's description β no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the βWhat this skill doesβ section above.
Prefer to read the source first? Open on GitHub.
When Claude uses it
Writes paid ad copy for Meta (Facebook and Instagram), Google Search, and LinkedIn: several variants, each testing a different hook, sized to each platform's limits. Use whenever someone wants ads, ad headlines, PPC or paid social copy, or variants to A/B test, even for one quick ad.
What this skill does
Ad Copy Generator
Purpose
Turn one offer or piece of source content into several genuinely different paid-ad copy variants β not paraphrases of each other, but different hook angles (pain point, benefit, social proof, urgency, etc.) suitable for real A/B testing β sized to fit each target platform's character constraints.
Step-by-step process
Plugin root: if a path below still shows an unexpanded
CLAUDE_PLUGIN_ROOT placeholder (Claude Code fills it in; other agents
such as Codex don't), resolve it before using any of them: take the
output of printenv CLAUDE_PLUGIN_ROOT if it prints a path, otherwise
walk up from this file's folder to the first folder that contains
references/brand-voice.md, and if neither works, ask the user where
this plugin lives. Use that absolute path in every file path and shell
command below.
-
Check onboarding status. Read
${CLAUDE_PLUGIN_ROOT}/references/brand-voice.md. If it doesn't exist, or its first line is<!-- MARKETING-SKILL:UNCONFIGURED -->, pause and ask the user this plugin's 4 setup questions (priority task; content types to produce; target audience + tone; default output format β same as/marketing-skill:marketing-setup) before continuing, then save the answers into that file and flip the marker toCONFIGUREDwith today's date. Otherwise, read it for tone, audience, and banned words below. If the file is unconfigured but the user already said to skip setup or use the defaults, don't ask: use its placeholder defaults for this run, say so in one line, and leave the marker as it is. -
Confirm scope.
- The offer/product/source content (paste, file, URL, or "our
product" β check
README*/package.json/pyproject.tomlat the project root first if so, same ascontent-repurposer). - Which platform(s): Meta (Facebook/Instagram), Google Search (RSA), LinkedIn Ads β default to asking rather than guessing, since limits and tone differ a lot by platform.
- The landing page URL and CTA, if any β kept identical across variants unless the user wants CTA itself tested as a variable.
- How many variants per platform (default: 3, each a different hook angle).
- The offer/product/source content (paste, file, URL, or "our
product" β check
-
Draft variants, one distinct hook angle per variant (e.g. pain point, concrete benefit/outcome, social proof, urgency/scarcity β don't reuse the same angle twice in one batch). Never invent a statistic, customer count, or claim not present in the source content.
-
Fit platform constraints. Apply the character guidance in "Platform constraints" below and show the actual character count next to each field β don't just assert it fits.
-
Self-check before finalizing: scan every variant against the banned-words list, confirm the CTA/link is identical across variants (unless CTA is deliberately being tested), and confirm no two variants in the same batch use the same hook angle.
When to use this skill
Trigger on requests like:
- "Write ad copy for [product/offer]"
- "Give me Meta ad variants for..."
- "Google ad headlines for..."
- "A/B test copy for our LinkedIn ad"
Platform constraints
These are commonly-cited platform guidelines, not values this skill can verify live β character limits and truncation points shift over time, so confirm current specs in the platform's own ad-manager UI before finalizing a real campaign, especially close to a limit.
| Platform | Field | Guidance |
|---|---|---|
| Google Search (RSA) | Headline | β€30 characters, up to 15 per ad |
| Google Search (RSA) | Description | β€90 characters, up to 4 per ad |
| Meta (Facebook/Instagram) | Primary text | Full text allowed, but plan for truncation around ~125 characters on most placements |
| Meta (Facebook/Instagram) | Headline | ~27β40 characters before truncation |
| LinkedIn Ads | Intro text | ~150 characters before truncation on most placements |
| LinkedIn Ads | Headline | ~70 characters before truncation |
Output structure (required)
Use this exact section order, as Markdown ## headings, per platform
requested:
[Platform Name]
A table per variant: Variant | Hook Angle | Headline (chars) | Body/Primary Text (chars) | Description (chars, if applicable) | CTA. Follow the table
with one line per variant naming why that angle was chosen.
Formatting rules
- Every variant in a batch must use a distinct hook angle β flag it if fewer distinct angles are possible than variants requested, rather than silently repeating one.
- Show a character count for every length-constrained field.
- Never fabricate a statistic, review count, or claim absent from the source.
- Apply
references/brand-voice.mdbanned words and tone, adapted to each platform's norms (e.g. LinkedIn skews more formal than Meta). - Keep the CTA/link identical across variants unless the user explicitly wants CTA tested as its own variable.
Example output
Fictional example: 2 Meta variants for "one-click export."
### Meta (Facebook/Instagram)
| Variant | Hook Angle | Headline (chars) | Primary Text (chars) | CTA |
|---|---|---|---|---|
| A | Pain point | "Stop Exporting By Hand" (22) | "Every manual export is ten minutes you don't get back. One-click export replaces the whole process β pick a report, done." (121) | Learn More |
| B | Concrete benefit | "Export Reports In One Click" (27) | "No templates. No copy-pasting. Just pick a report and format β it's live for every customer today." (98) | Learn More |
- Variant A leads with the friction the buyer already feels (manual
export as wasted time).
- Variant B leads with the mechanism itself as the selling point (how
fast and simple it now is).
Related skills
Apple Ads Analyzer
AgriciDaniel
Analyze Apple Ads campaigns for structure, bidding, attribution, and iOS app performance.
Attribution Health Audit
AgriciDaniel
Audit your cross-platform attribution setup for iOS, GA4, consent compliance, and conversion accuracy.
Ad Format Resizer
SupercmoHQ
Resize image ads to different dimensions while preserving product, copy, and design.
ckm:banner-design
nextlevelbuilder
Generate social media banners, ads, and website headers in multiple styles.