← Both skills · static-ads · SKILL.md
Make winning static ads for ANY brand, from analysis to finished files, with a review loop that never regresses. Everything needed is inside: find the angles (product → mechanism → angle, avatar, research winners, decode templates), clone winning templates into the brand's own ads (spec, checker, image engines, real product pasted by script, full-size QC), review 10 at a time by Loom, and turn every comment into a permanent rule. The only input it needs from you is the brand kit (voice, facts, offer, product photo). Use for: "make ads for <brand>", "clone these templates", "new batch", "analyze this ad / competitor", "find the angles", "here is my Loom", "redo the ads", "fais-moi des ads", "nouveau batch", "trouve les angles". For Wildhorn, use wildhorn-static-ads (same skill, brand already loaded).
What it does: takes ads that already win (templates) and turns them into the brand's own ads with its real product, then learns from every review so the next batch is better.
Proven: on Wildhorn, the first batch had 0 % of ads accepted on the first try. After this system: 71 % first try, 87 % after one fix, 100 % final, 1.4 review rounds per ad instead of 2.5.
The only thing you bring: the brand kit (BRAND-KIT.md + brand.json) and the templates to clone.
| # | Step | What happens | Open |
|---|---|---|---|
| 1 | Brand kit | Fill the brand's voice, facts, offer, proof, avatar, banned words, product photo. Never invent a fact: unknown = [MISSING], ask. |
BRAND-KIT.md, brand.json |
| 2 | Understand | Product → mechanisms → angles. Who buys (avatar). Find ads that win (research). Decode each template: what it says, its visual trick, how it was made. | 1-understand/ |
| 3 | Make | One spec per template (what to KEEP, what to SWAP). The checker blocks bad specs. Generate, paste the real product + exact text by script, check every ad at full size. | 2-make/ |
| 4 | Review | 10 ads per batch on a review page. The founder records a Loom. Transcribe it, write each verdict. | 2-make/review-and-pages.md |
| 5 | Improve | Every comment becomes a permanent rule (and an automatic check when possible). Fix the misses, keep the wins frozen. | 3-improve/ |
| 6 | Measure | First-try rate, after-one-fix rate, final rate, rounds per ad. Compare with the last batch. | 3-improve/stats.md |
Then repeat from step 3 with the next 10.
brand.json. Never invent a number.The full rulebook (97 rules + prompt craft + system rules, each with the case it came from):
3-improve/rules.md.
| You need to… | Open |
|---|---|
| Start a new brand from zero | BRAND-KIT.md, then 1-understand/new-brand-worksheet.md |
| Find mechanisms and angles | 1-understand/product-mechanism-angle.md, 1-understand/avatar-angle-system.md |
| Find winning ads to clone | 1-understand/research-playbook.md, 1-understand/competitor-case-studies.md |
| Understand why an ad works | 1-understand/ad-dna.md, 1-understand/frameworks-evolve-crafter.md, 1-understand/copy-analysis.md |
| Decode a template before cloning | 1-understand/template-reading.md |
| See the analysis workflows in full | 1-understand/WORKFLOWS.md |
| Write a spec | 2-make/spec-format.md |
| Pick the image engine / handle a refusal | 2-make/engines.md |
| Paste the product and text | 2-make/composite.md |
| Check an ad before showing it | 2-make/qc-checklist.md |
| See the full production pipeline | 2-make/PIPELINE.md |
| Turn a Loom into rules | 3-improve/feedback-loop.md |
| All the rules | 3-improve/rules.md |
| Compute the stats | 3-improve/stats.md |
| Run the scripts | scripts/README.md |