Analyze competitor ads from sources such as Meta Ad Library and TikTok Creative Center to identify patterns in hooks, formats, offers, and proof. Generate original concepts for your product using claims you can support, check them for similarity, and pass them to production skills.
Turns winning competitor ads from the Meta Ad Library, TikTok Creative Center, Kalodata or FastMoss into new ad angles for your own product, without copying them: rank ads by winner signals (running 30+ days, many variants, reach), break the top ones down with varg video_analyze (hook, angle, format, offer, proof, CTA, awareness level), map which angles and formats the market already runs, generate 10-20 new concepts (same angle in a new format, same format with a new angle, one awareness level up or down) using only claims you can prove, check originality, and hand each concept to a varg skill to produce. Use when a seller says "clone this winning ad", "what are my competitors running", "find winning ads for my product", "spy on TikTok Shop ads", "give me new angles", or pastes Ad Library or Creative Center links. Not for re-uploading or re-cutting someone else's footage (never), and not for one ad's 20 variants (use varg-ad-variants).
Run score Not tested
See how to installReview the instructions the agent will receive before you install.
Sellers' top playbook in 2026: find ads that have run for 30+ days and remake what made them work. The idea, the format and the structure are free to reuse; the footage, the copy, the music, the voice, the logo and a beat-for-beat storyline are not (idea vs expression). This skill keeps you on the right side of that line and turns winners into concepts for your product.
| Source | How | What you get |
|---|---|---|
| Meta Ad Library | the person pastes URLs (search the competitor page or a keyword) | start date, variants, landing page; EU/UK ads also show reach |
| TikTok Creative Center → Top Ads | the person pastes URLs | likes, CTR tier, budget level, by industry and region |
| Kalodata / FastMoss | exports | TikTok Shop products and the videos that sold them (GMV) |
| Ad Library MCP scrapers | facebook-ads-library-mcp, tiktok-ads-library-mcp | started_running, ads_using_creative |
The official Meta ads_archive API returns non-EU ads only if they are political or issue ads, so it doesn't work for product ads (Meta docs). Prefer pasted URLs or a scraper; outside the EU/UK an ad's data vanishes when it pauses.
Winner signals (heuristics, not Meta facts): running 30+ days = proven, 14–30 = gaining; 5+ variants of one concept or the same video across pages = being scaled; EU reach = the only real volume number.
Download with media-download (vdl <url>), upload (POST /v2/files or create_upload), then:
curl -s -X POST https://api.varg.ai/v2/tools/video_analyze/call -H "Authorization: Bearer $VARG_API_KEY" \
-H "Content-Type: application/json" -d '{"model":"video-analyze","video_url":"AD_URL",
"prompt":"Break this competitor product ad down for a remix: every shot with start-end seconds, camera and what happens; all on-screen text with timings; audio; the hook in the first 3 seconds; the angle (which pain or desire it sells), the format, the offer, the proof moment and the CTA; and the awareness level of the viewer it targets."}'
About 5 credits each. Tag each ad in one row of winners.json:
{ "url": "…", "days_running": 41, "variants": 6, "hook": "spiking a drink at a club bar", "angle": "safety disguised as cute nail art",
"format": "problem → Vs → solution, music, text overlays", "offer": "20 stickers", "proof": "sticker turns green in the drink",
"cta": "link in bio", "awareness": "problem-aware" }
Put the winners in a grid of angle (pain, desire, fear, identity, price, gift, comparison) × format (problem → solution, unboxing, demo, testimonial, before/after, comparison, tutorial, POV skit, song/mascot) × awareness (unaware, problem-aware, solution-aware, product-aware, most-aware). Empty cells next to proven ones are the opportunities (Motion frameworks).
Generate 10–20 concepts with three moves:
Each concept: title, angle and why it works, format, hook line (≤3 s), proof shot, claims allowed (from the seller's evidence only), the varg skill that makes it.
Reject a concept if it reuses footage, music, voice, copy lines, logos or characters, or repeats the source storyline beat for beat with only the product swapped. Then run every concept through ai-ad-compliance-check before anything is generated.
Hand each concept to the skill that makes it: varg-product-hook (hooks, problem → Vs → solution), varg-unboxing-video, ai-ugc-ad-video, varg-mascot-jingle-ad, varg-dropship-reel, then varg-ad-variants for hook variants of the winner.
2026-10-28. Source: a German drink-test nail sticker reel (club bar → CCTV corridor → "Vs…" → product → sticker turns green → link in bio). video_analyze (5 credits) returned 7 timed shots, all text, angle "safety disguised as cute nail art", format problem/solution, proof "sticker turns green", awareness problem-aware. Remix: same format, new angle for a different product, car seat gap fillers: a dashcam "problem" (phone falls into the seat gap while driving), "Vs…", the reveal, the proof (coins and a phone stay on top), "Link in bio". Nothing reused but the structure. Result (21 s): https://s3.varg.ai/files/194b94d0-58ff-4e8b-91de-4f1959d04013/file_a2aa0853b683.mp4 · Steps: references/example.md.
varg-product-hook (breakdown steps in references/breakdown.md), ai-ad-compliance-check, varg-ad-variants, ad-creative-testing (test plan), media-download.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Plan worthwhile local tasks by checking dependencies, effects, resources, and platform boundaries, then execute the exact approved plan. It can use safe parallel work or suitable Apple-silicon backends when the project exposes them.
Turn a winning TikTok or Meta ad into a structured creative test plan. Identify what made it work, plan distinct concepts, set test budgets and review timing, and use funnel metrics to decide whether to stop, iterate, or scale each ad.
Record deferred engineering decisions and temporary workarounds in a debt registry, with context such as effort, ongoing cost, affected area, and payoff trigger. Entries receive short labels so they can be reviewed or dropped later.
Generate or compare AI agent code for LangChain, EINO, AutoGen, AgentScope, and CrewAI using framework-specific templates, API mappings, conventions, and project structures. Reference files include verification dates to help identify potentially outdated guidance.
Investigate a reproducible bug, test failure, or unexpected page behavior, then make a narrowly scoped fix. The workflow calls for a regression test before changing code, verifies the fix, and reports remaining risks and evidence while avoiding unrelated refactoring or feature work.
Build an AI agent project through six collaborative stages, from requirements and architecture to tools, prompts, code generation, and verification. Start framework-independent, then generate a runnable project for frameworks such as LangChain, EINO, AutoGen, AgentScope, or CrewAI.