find-winning-ads
Skill0 starsChanged 27 days agoMITView on GitHub
Find winning Facebook and Instagram ads in the AdSwipeFile ad library by niche, format, brand or keyword, and present them with links. Use for "show me winning ads in {niche}", "what's working in {niche} right now", ad inspiration, ad examples, or examples of a format (UGC, before and after, testimonial, comparison).
In this repository14 other files it gives its agents.
- audit-ad-creativeSkill
- competitor-adsSkill
- creative-strategySkill
- generate-ad-conceptsSkill
- get-startedSkill
- map-competitorsSkill
- map-personasSkill
- pick-creative-mechanicSkill
- record-creative-learningSkill
- review-saved-adsSkill
- teach-ad-breakdownSkill
- track-competitor-adsSkill
- weekly-competitor-reviewSkill
- write-hooks-from-adsSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Call each list tool at most once per conversation.
- Very specific: line 42
Call list_ads with the slugs, and a limit close to what the user asked for.
- Distinctive: line 49
Never call it a launch or run date.
- Distinctive: line 50
If fewer ads match than asked, say how many there are.
- Very specific: line 52
Use nextCursor only when the user asks for more.
- Very specific: line 56
Only read an ad in full with get_ad when the user wants its copy, or when the task needs the copy (see the write-hooks-from-ads skill).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Find winning ads
Needs the AdSwipeFile connector. This skill reads the AdSwipeFile ad library of Meta ads through its MCP server. If no AdSwipeFile tools (
search_library,list_ads) are available, say so before anything else and give the step for this host. Claude Code:claude mcp add --transport http adswipefile https://adswipefile.com/api/mcp. claude.ai or ChatGPT: addhttps://adswipefile.com/api/mcpas a custom connector. Other clients: https://adswipefile.com/docs/mcp?utm_source=agent-skill&utm_medium=referral&utm_campaign=find-winning-ads. Connecting opens a sign-in to an AdSwipeFile account. Steps that need only the user's own material can go on; never invent ads, brands or links in place of the library.
Resolve names to slugs first
list_ads filters on slugs. Turn the user's words into them before listing:
- A product, idea or loose phrase ("protein powder", "sleep gummies"): call
search_libraryand use the brands, niches or ads it returns. - An industry ("skincare", "SaaS"): match it against
list_niches. Niches are broad (skincare sits under Beauty / Personal Care), so pick the closest and say which one you used. - A format ("UGC", "before/after", "reviews"): match it against
list_visual_formatsand filterlist_adsbyvisual_format. When the user asks what makes an ad that format, or which format an ad is, callget_formatwith the slug: it says what must hold, the near misses and the formats it is mixed up with. - A move or a stage ("objection handling", "cold traffic"): filter
list_adsbymechanic(the slugs in pick-creative-mechanic) orawareness(unawaretomost-aware). - A brand: call
search_librarywith its name. If no brand matches, say the library does not have it rather than substituting another brand, and offer to report it withsubmit_feedback.
Call each list tool at most once per conversation; reuse what it returned.
List and present
- Call
list_adswith the slugs, and alimitclose to what the user asked for. Results are the newest additions to the library first. For "what's working" or "what's been running longest", passsort: "longest_running"; a long run is a reason to study an ad, not proof of its results. - For each ad give the brand, niche, formats and its page
url. Show theimagewhen the host renders images. created_atis when the ad was added to the library, not when it ran. Never call it a launch or run date.- If fewer ads match than asked, say how many there are. Offer to widen the search (drop the format, or try a neighbouring niche) instead of padding.
- Use
nextCursoronly when the user asks for more. When the list answers with anotethat it was cut short, pass it on once with its link and narrow with filters instead of paging.
Only read an ad in full with get_ad when the user wants its copy, or when the
task needs the copy (see the write-hooks-from-ads skill).
Keep the ones worth keeping
Offer, once, to keep the ads the user likes: save_ad, or a board named for
the search (create_board, add_ad_to_board), with a set_ad_note on why
each is worth studying. Only on a yes.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

