Rapidly generate up to 25 distinct ad creative hypotheses for testing. Produces a matrix of angles (pain, benefit, social proof, urgency, curiosity, comparison) x formats (static, carousel hook, UGC-style, bold-text overlay) with headline, visual prompt, and rationale for each, then generates the visuals with any image tool.
SkillScore82Maintained
Automatic
◆Marketing & Content pack
Last verified 8/6/2026
What it changes
Passed 17 of 21 test cases with the skill, and 8 of 21 without it.
Without skill36%
With skill82%
Checks it passed that the model alone missed
Outputs 12 distinct ad hypotheses.
Caps the total generated hypothesis count at 25.
Restricts all generated headlines to 7 words maximum.
and 10 more in the full report
10 fixed8 passed either way4 still failing0 broke
Pass-rate lift
+45 pp
Tokens
+36%
Agent turns
0%
Response time
+0%
Model gemini-3.6-flashMeasured 8/6/2026v118 of 22 cases passed with skill · 21 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 9
Prompt
We want to test ads for a new AI coding assistant on Meta feed. What key audience detail do you need from us before generating the ad hypothesis matrix?
What was checked, without and with the skill
Checked
Without skill
With skill
Requests the audience primary pain point or desire before generating hypotheses.
Missed
Passed
case-16 · gemini-3.6-flash · 8/6/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md64 linessingle file
Creative — Ads Explorer
Pump out a broad, diverse batch of ad hypotheses for A/B/n testing. This is the exploration stage: quantity + diversity beat polish. The goal is to cover the message space so the market tells you what works — not to perfect one ad.
When to use
Launching paid social/display and need a test batch.
An existing ad is fatiguing and you need fresh angles.
Validating which message resonates before investing in production.
When NOT to use
You already know the winning angle and just need it finished → creative-polish.
You're merchandising a specific discount/bundle → creative-offer (sharper for that).
No visual direction exists and brand look matters → lock it via creative-explore first.
Inputs (ask only what's missing)
Product + core offer (what they get).
Audience + their #1 pain or desire (the emotional hook).
Formats: static hero, hook-frame (thumb-stopper top third), UGC mock (handheld, casual), bold-text overlay, before/after split, problem→solution 2-panel.
Not every cell is worth filling — pick the combinations most likely to resonate for THIS audience.
Workflow
Pick the count (default 12, max 25). Choose the highest-potential angle×format cells.
For each: headline (≤7 words), subtext, visual prompt, and why it might win (the hypothesis being tested).
Kill near-duplicates aggressively — two ads testing the same idea waste spend.
Generate visuals with your image-generation tool at platform ratio (1024x1024 square, 1792x1024 landscape, or vertical for stories/TikTok). Save to ./creative/ads/<NN>-<angle>.png.
Output a ranked table; mark the top 3-5 to launch first and state what each one is testing.
Worked example
Product: AI meeting-notes app. Audience pain: "I forget action items." Platform: Meta feed (1:1).
| # | angle | format | headline | tests hypothesis | pri |
|----|------------|--------------|----------------------------|-------------------------------|------|
| 1 | pain | hook-frame | "Forgot the action item again?" | pain-first beats feature | HIGH |