agentleFS
adswipefileadswipefile/adswipefile-plugin/skills/generate-ad-concepts/SKILL.md

generate-ad-concepts

Skill0 starsChanged 27 days agoMITView on GitHub

Generate ad concepts by walking the map from anchor, persona, awareness stage and trigger moment to angle, offer statement, mechanic, format and hook, one concept or a full-funnel batch, with reference ads from the AdSwipeFile ad library for each step. Use when asked for ad ideas, ad concepts, angles, creative briefs, a creative roadmap, funnel coverage, or to turn a format, hook or angle someone likes into a concept.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Ask once for the minimum missing: the product, who buys it, and at least a few customer words.

    line 32
  2. Distinctive:

    Funnel order: Start at product aware, the people choosing now.

    line 56
  3. Very specific:

    Read a sample with get_ads (up to 25 per call) for breakdown and why_it_works.

    line 70
  4. Very specific:

    Before a concept ships: audit-ad-creative.

    line 103
  5. Distinctive:

    Never state a reference ad's results: spend, ROAS, conversions, or that one ad beat another.

    line 107
  6. Distinctive:

    Write "kept running for 94 days".

    line 108

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Generate ad concepts

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: add https://adswipefile.com/api/mcp as a custom connector. Other clients: https://adswipefile.com/docs/mcp?utm_source=agent-skill&utm_medium=referral&utm_campaign=generate-ad-concepts. 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.

One pass through the map makes one concept; many passes cover a funnel. A concept is not finished until it has a value at every step, and every value comes from the user's research and results or from ads in the library, never from this file. Words (anchor, persona, trigger moment, angle, offer statement, mechanic, funnel job) are used as the creative-strategy skill's glossary defines them.

0. Load what the user gives

The user's own material comes from the user: pasted, uploaded, linked, or from another connector they have. Take what is there:

  • Product facts, the offer, proof points (a claim missing from them can't be made), voice, don'ts.
  • Customer words: reviews, survey answers, personas (map-personas), verbatim quotes.
  • Past results: which of their ads ran, what they learned, concepts already made.

Ask once for the minimum missing: the product, who buys it, and at least a few customer words. Thin research: say so, work from what exists, and mark each concept "unvalidated".

The map

  1. Anchor one pain or desire, from the customer words, that the product credibly solves. Two pains pulling at you are two passes.
  2. Persona sharing that anchor and a life context. Deep enough to write from inside their day.
  3. Awareness stage where the ad finds them. Each stage has a job: make the problem exist (unaware), name and sharpen it (problem aware), position the category (solution aware), win the comparison (product aware), close (most aware). Pick a stage that is true.
  4. Trigger moment that moves them one stage on, from the personas or the customer words. Filmable: a where, a when, what happened. Different moments are different concepts.
  5. Angle: the one product truth that answers what the moment made urgent. If it has to bend to fit the stage or the moment, change the angle or the moment.
  6. Offer statement: one or two sentences, in the persona's words, of what the product does for them in that moment. Names the person and the now; otherwise it is a tagline.
  7. Mechanic: how the viewer arrives at the point. See pick-creative-mechanic.
  8. Format: the container that carries the offer and stages the moment. list_visual_formats for the vocabulary, get_format for how to build one.
  9. Hook: the opening line and first frame together, usually the first second of the trigger moment. Write several in the customer's words and keep the one this persona stops for. Tactics: list_hooks.

Funnel order. Start at product aware, the people choosing now. Most aware is for deliberate pushes (a sale, a deadline). Once product aware has a concept the user's results back, climb: solution aware, problem aware, unaware, where new customers come from.

References from AdSwipeFile

For a step you can't fill, study how other brands built it. Keep calls few; before a batch, get_usage says how many are left.

  • list_ads with the concept's awareness stage and the step you need: mechanic (the slugs in pick-creative-mechanic), visual_format or hook, plus a niche, and sort: "longest_running". Two tag filters at most; drop one when nothing matches.
  • search_library with 1 to 3 plain words for a moment or a move ("text message", "before after").
  • A competitor: list_ads with its brand for brand_patterns.
  • Read a sample with get_ads (up to 25 per call) for breakdown and why_it_works.

A reference shows a construction. Its long run is a reason to study it, never proof the concept will work. Cite every reference by its url.

Entering mid-map

A format, hook, angle, persona or trigger someone already likes enters at its step. Walk the rest explicitly: name the persona, stage and moment a hook implies, the stage a format serves, the moment an angle answers. A step you can't name means it is a clever line, not a concept: give it in the reply as a hook idea and stop there.

Each concept as a brief

In the reply (or a file, if the client can write files), one brief per concept:

  • Title: persona + trigger moment, short.
  • Offer statement.
  • One line per step: anchor, persona, stage, trigger moment, angle, reason to believe, mechanic, format, funnel job, and why it should land, citing the customer words it rests on.
  • Hook options: three or more, in the customer's words.
  • Opening scene and format.
  • Reference ads, by url, each with the step it shows.

Then offer to file the references on AdSwipeFile: create_board (or an existing one from list_boards), add_ad_to_board, and set_ad_note saying which step each ad shows, in the client's workspace when the user works for several (list_workspaces). Only on a yes. Once it is filed, offer share_board so the user can send the references to a client or team as one link.

Batches: one brief per concept, then the coverage the batch leaves open (personas with no upper funnel concept, stages with one trigger moment, angles never briefed). Call the set a roadmap.

Done when every concept has all nine steps, each brief is in the reply, and the gaps are named. Before a concept ships: audit-ad-creative.

Rules

  • Never state a reference ad's results: spend, ROAS, conversions, or that one ad beat another. Write "kept running for 94 days".
  • Numbers and claims in a concept come from the user's proof points, never from a reference ad.
  • Competitor reviews teach the category's words; they are never this brand's proof.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.