Skip to content

mbfinotti/advertising-skills / skills/ad-format-fit

ad-format-fit

Verified

Flag ad formats that are a poor fit or commonly misused for a given objective and platform - formats the objective does not even expose, formats built for a different funnel stage, formats the team cannot produce - then, post-launch, separate a format problem from a creative or targeting problem. Use whenever the user asks which ad format to use, mentions carousel, Stories, in-feed vs in-stream video, a funnel-stage mismatch, a pre-launch format check, or says a format 'isn't working' - even if they never say 'format fit'. Covers B2B and B2C, with explicit launch blockers. Do NOT use to choose the channel itself (mbfinotti/advertising-skills@ad-platform-selection) or to brief the creative (mbfinotti/advertising-skills@ad-creative-brief).

FileKindSize
SKILL.mdskill md34.8 KB
evals/evals.jsonconfig22.8 KB

evals/evals.json

View on GitHub
{
  "skill_name": "ad-format-fit",
  "evals": [
    {
      "id": 1,
      "prompt": "I run growth at Fienzo, a B2B workflow-automation SaaS. First paid social push: a conversions campaign optimized for demo-request submissions at $25/day, target CPA around $80. Creative plan is a 5-card carousel walking through our feature set, aimed at cold lookalike audiences. Right now we get about 2-3 demo requests a week across all channels. Can you sanity-check the format setup before I build it?",
      "expected_output": "A format-fit report that rules the conversions objective out on the conversion-volume floor, computes the budget floor from the target CPA, redirects to a traffic objective with manual measurement, flags the carousel as wrong for cold first touch with a substitute, and delivers a dated per-format verdict table.",
      "files": [],
      "expectations": [
        "States that conversion-optimised delivery needs roughly 50 optimisation events per ad set per rolling 7 days",
        "Concludes the proposed campaign cannot clear that volume floor using the user's own figures ($25/day, ~$80 target CPA, 2-3 demo requests per week)",
        "Derives the daily budget floor as (target CPA x 50) / 7, roughly $570/day at an $80 CPA",
        "Recommends a traffic or website-visits objective with manual funnel measurement instead of forcing the conversions objective",
        "States that under-volume conversion optimisation is not neutral: the campaign pays more per click for a system that cannot learn",
        "Flags the 5-card carousel as structurally wrong for cold first touch",
        "Gives the structural reason: a carousel demands sequential swipe attention a cold audience does not give",
        "Recommends single image (or another eligible near-zero-effort format) as the substitute for cold prospecting",
        "Distinguishes hard gates that block a launch from soft practitioner preferences that never do",
        "Delivers a per-format verdict table using fit / conditional / misfit / ineligible verdicts with a specific reason per row",
        "Labels the platform thresholds cited as an August 2026 snapshot or as unverified-current, and tells the user to confirm them in the ads interface before launch",
        "Marks the conversions objective as conditional (fit once weekly conversion volume clears the ~50-event floor) rather than banning it permanently"
      ]
    },
    {
      "id": 2,
      "prompt": "I'm head of demand gen at Kordelia Systems, a B2B data-governance platform. Q4 plan on LinkedIn targeting compliance leaders in Germany, France and the Netherlands, all cold: (1) sponsored inbox messages offering a demo, (2) promoting our CEO's posts as thought-leader ads with a Book a demo button, (3) a connected TV test for credibility. Total paid budget is $9K/month, and our site retargeting pool there is about 210 members. Which of these formats should we run?",
      "expected_output": "A verdict that rules all three formats ineligible or misfit for this plan - inbox blocked by EU consent for cold audiences, thought-leader ads locked out of direct response with no CTA button, connected TV awareness-only and below its spend threshold - plus the audience-floor failure on the 210-member pool and substitute feed/document formats.",
      "files": [],
      "expectations": [
        "Flags inbox/message ads to cold EU audiences as reaching opted-in members only",
        "Classifies the EU inbox restriction as a legal/consent hard gate, not a performance preference",
        "Moves the cold EU outreach to feed and/or document-style formats as the substitute",
        "States thought-leader ads run only under awareness, engagement, and video-view objectives",
        "States thought-leader image and video variants carry no call-to-action button, so a Book-a-demo direct-response unit is structurally impossible with them",
        "Flags connected TV as an awareness-only format unfit for demo generation (non-clickable, lean-back viewing)",
        "States connected TV on the professional network is generally only worth funding above roughly $15K/month total spend, so at $9K/month it cannibalizes budget from measurable formats",
        "Mentions at least one further connected TV constraint: automated bidding only, a daily minimum around $50, or limited geographies",
        "Flags the 210-member retargeting pool as below the 300-member minimum, so campaigns against it will not serve",
        "Cites the roughly 300,000 audience floor for sponsored content and inbox formats (or the ~50,000 general working floor) when sizing the cold audiences",
        "Marks each blocked format as ineligible with the specific gate named, instead of ranking it low",
        "Recommends a substitute format for every non-fit row",
        "Attaches a date and verification status (August 2026 snapshot or unverified-current) to the platform numbers relied on",
        "Does not include the sponsored inbox messages to the cold German, French, or Dutch audiences in any recommended plan"
      ]
    },
    {
      "id": 3,
      "prompt": "One-person marketing team at Bruntveld, a DTC kitchenware brand. Our old agency left us a 2-minute 16:9 brand film (dialogue-driven, beautiful) and a folder of product statics. I can make about two new assets a month, tops. For our anniversary sale launching in 10 days I want Reels/Stories full-screen vertical plus feed placements, and everyone says creator-style UGC video is what actually works. How should I set up the formats?",
      "expected_output": "A capacity-gated plan that excludes vertical placements rather than letterboxing the 16:9 film, captions it for muted feeds, re-cuts the film instead of running it whole, drops UGC for this deadline while naming the deadline as the reason, and matches placements to a two-asset-per-month refresh reality.",
      "files": [],
      "expectations": [
        "Excludes full-screen vertical placements at setup instead of letting automatic placement letterbox the 16:9 film",
        "States the reason: letterboxed landscape in a vertical slot reads instantly as repurposed",
        "Flags the dialogue-driven film for silent-autoplay feed placements and requires full captions, citing that roughly 85% of feed video is watched muted",
        "Applies format lifespans: statics roughly 14-28 days, short video/carousels 21-35, creator-style 28-42",
        "Concludes two assets per month cannot feed a placement that burns creative in about three weeks",
        "States the 10-day deadline promotes near-zero-effort formats and demotes anything needing a shoot, naming the deadline as the answer that moved the ranking",
        "Does not recommend commissioning creator/UGC video for this launch, and says the short deadline is why that rung is unreachable",
        "Flags running the 2-minute film as the primary ad as a misuse: completion collapses on long-form primary ads (documented cases under 10%)",
        "Fixes the film at the re-cut rung - chop it into short videos plus statics - rather than commissioning new assets",
        "Picks substitute formats by efficiency inside the objective, not by cheapness",
        "Computes or applies standing load as effort per asset multiplied by refresh rate",
        "Requires a distinct cut per placement, or narrows the placement list to what one person can version",
        "Labels capacity-blocked formats as production-capacity misfits in the verdict rather than merely ranking them lower",
        "Presents the ranking as a default that would reverse under a compounding mandate (a shoot buys footage and rights that make later cuts an editing job), not as a law"
      ]
    },
    {
      "id": 4,
      "prompt": "Meraluna Skincare here. Our short vertical video campaign: 3-second view rate 31%, hold to 15 seconds 4%, CTR 0.3%, purchases near zero. Meanwhile a static-image campaign converts fine. Separately, our older hero video that crushed it for three weeks is now fading (day 26, frequency 2.1 on cold prospecting). The team's conclusion is vertical video just doesn't work for our brand and we should move everything to carousels. Are they right?",
      "expected_output": "A diagnosis that refuses the format verdict: the strong-thumbstop/collapsed-hold pattern is a clickbait opening (a creative problem), the day-26 decline is ordinary fatigue within the format's lifespan, frequency is in band, dashboard cross-format comparison is contaminated, and confirmation via placement breakdown must precede any restructure.",
      "files": [],
      "expectations": [
        "Rejects the immediate format verdict and treats the suspected format failure as a hypothesis to confirm",
        "Applies the refresh priority order hook, visual treatment, format, then body copy/CTA - placing format third, not first",
        "Localises the failure top-down through the delivery funnel stages: stop (thumbstop/3-second view), stay (hold), click (CTR), convert (post-click conversion)",
        "Diagnoses the 31% 3-second view with 4% hold as a clickbait opening / on-ramp problem in seconds 3-15 of the creative, not a format problem",
        "Routes the opening problem to hook-level analysis rather than a format swap",
        "Diagnoses the day-26 decline as fatigue arriving within the short-video expected lifespan of 21-35 days",
        "Hands the fading hero video to a dedicated creative wear-out diagnosis instead of relitigating the format",
        "Notes frequency 2.1 on cold prospecting sits inside the comfortable ~1.0-2.5 band, so it does not implicate audience size or placement supply",
        "Warns the video-versus-static dashboard comparison is contaminated: delivery optimises for marginal, not average, cost, and the delivery system chose who saw which asset",
        "Orders confirmation methods placement-and-format breakdown first, then an isolated ad set, with a holdout last",
        "States the escalation rule: escalate one rung only when the previous method stays ambiguous, and reserve the holdout for a decision moving a whole budget line",
        "Refuses to move everything to carousels on current evidence",
        "Names at least one signal that would genuinely implicate the format (placement-breakdown divergence, or failure persisting across every creative variant in one surface)"
      ]
    },
    {
      "id": 5,
      "prompt": "We got a consultant deck at Vantorra Legal, a B2B contract-review SaaS. It says: (a) move demo bookings from our landing page to in-platform lead forms since forms convert at 13% vs 4% for landing pages; (b) run 6-second bumper ads for demo requests because human attention span is 8 seconds, shorter than a goldfish's; (c) shift most budget to carousels since they deliver 30-50% lower cost per conversion. Our landing page converts at 6.5% and has an embedded booking calendar. Budget isn't the constraint. Which recommendations do we implement?",
      "expected_output": "A rejection of all three recommendations as stated: the form-vs-page stat conflates studies and the measured 6.5% page with embedded calendar keeps demo traffic, the goldfish claim is discarded outright and bumpers are wrong for direct response, and the carousel cost claim has no reliable independent support.",
      "files": [],
      "expectations": [
        "Rejects the 13% vs 4% form-versus-landing-page comparison as conflating different studies and audiences",
        "States qualified-opportunity rate frequently runs the other way because form friction filters for intent, and judges lead formats on downstream qualified-opportunity rate rather than cost per lead",
        "Keeps demo booking on the landing page: demo/trial offers get a page, because a form adds friction to a calendar-booking goal",
        "Applies the measured threshold: a landing page converting at 5%+ (here 6.5%, with an embedded calendar) receives the traffic - a measured conversion rate decides, not taste",
        "Discards the 8-second/goldfish attention-span claim entirely as not describing real user behavior",
        "Flags 6-second bumpers as structurally wrong for direct-response demo requests: a forced few seconds cannot carry a considered purchase",
        "If bumpers survive at all, restricts them to awareness and points direct-response interaction at click-priced in-feed video",
        "Rejects the carousel 30-50% lower cost-per-conversion claim as unsupported by reliable independent evidence",
        "Positions carousel as fitting consideration, multi-product, warm audiences rather than an all-in format",
        "Applies the counterweight: no format wins every time (video wins only around 60% of head-to-heads), so every per-format verdict is conditional, never absolute",
        "Reserves in-platform lead forms for content or webinar registration, or for a landing page converting under roughly 2% - not for demo booking",
        "Names the qualitative test: collecting a lead versus starting a sales conversation",
        "Recommends implementing none of the consultant's three claims as stated"
      ]
    },
    {
      "id": 6,
      "prompt": "Pellonia Home, new ecommerce brand, brand-new ad account, first campaign ever. Our freelancer's plan: the fully-automated multi-surface shopping campaign type at $40/day with our 8 products and 6 creatives. We also booked a retail-media display buy - our 970x250 banner was approved at upload so that's sorted, but that campaign is barely spending and he wants to raise bids. And our B2B sister brand wants a LinkedIn lead form with 14 qualifying fields. Pixel was installed last week, never tested. Run the pre-launch check.",
      "expected_output": "A pre-launch QA report that blocks the automated multi-surface type as a first campaign on a new account with unverified tracking, treats the product/creative counts as flags not blockers, names the silent-exclusion trap behind the underdelivering display buy, blocks the 14-field form on the 12-field maximum, and states the zero-blocker launch threshold.",
      "files": [],
      "expectations": [
        "Advises against launching the fully-automated multi-surface shopping type as the account's first campaign",
        "Names the disqualifying conditions present: below the roughly $50/day comfort floor, a new account, and weak/unverified tracking",
        "Recommends starting with standard search/shopping campaigns and graduating later",
        "Treats the 30+ products and 15+ creatives readiness numbers as practitioner defaults to flag, not platform minimums - does not declare 8 products or 6 creatives ineligible on those grounds",
        "Marks the never-verified pixel as a launch blocker requiring the conversion event to show recent, verified firing (browser-side and server-side) before spend",
        "Names the silent-exclusion trap: an asset can be approved at upload yet silently excluded from mismatched placements",
        "States that upload acceptance proves neither policy approval nor delivery",
        "Re-diagnoses the barely-spending display campaign as a likely placement/dimension exclusion masquerading as a budget or bid problem, and advises against raising bids as the fix",
        "Flags the 14-field lead form against the 12-field maximum on the professional network",
        "Recommends 3-4 form fields and CRM field mapping for the lead form",
        "States the launch threshold explicitly: any single failing blocker item means do not launch (zero blockers is the threshold)",
        "Delivers a launch-blocker list with a fix attached to each failing item",
        "Includes previewing the ad in every surface it will serve (feed, vertical, in-stream, partner network) among the blockers, checking truncation, cropping, and call-to-action",
        "Applies vertical safe zones - key elements out of roughly the top 14% and bottom 20-35% - or equivalent spec enforcement via export presets and a spec-bearing filename convention"
      ]
    }
  ],
  "trigger_queries": [
    { "query": "Should we use carousel or single image for our cold prospecting campaign?", "should_trigger": true },
    { "query": "Which ad format works best for lead gen on LinkedIn?", "should_trigger": true },
    { "query": "Is a Stories placement right for a B2B demo campaign?", "should_trigger": true },
    { "query": "in-feed vs in-stream video for our product launch?", "should_trigger": true },
    { "query": "We want to run connected TV ads for demo requests, good idea?", "should_trigger": true },
    { "query": "Our carousel isn't working, should we switch to video?", "should_trigger": true },
    { "query": "Pre-launch format check for our Q4 campaign please", "should_trigger": true },
    { "query": "Can I run a 2-minute brand film as our main paid ad?", "should_trigger": true },
    { "query": "Do bumper ads make sense for driving purchases?", "should_trigger": true },
    { "query": "Lead form or landing page for our webinar signups?", "should_trigger": true },
    { "query": "The client wants Stories ads but we only have landscape video, what do we do?", "should_trigger": true },
    { "query": "Is vertical video worth it if we sell B2B software?", "should_trigger": true },
    { "query": "Which formats can I even run under an awareness objective?", "should_trigger": true },
    { "query": "Why does our ad look cropped in some placements?", "should_trigger": true },
    { "query": "Are message ads a good idea for cold outreach in Germany?", "should_trigger": true },
    { "query": "My video ad gets views but no sales, is video the wrong format?", "should_trigger": true },
    { "query": "What ad types should a tiny team with no video capability run?", "should_trigger": true },
    { "query": "Format recommendations for retargeting warm traffic on social", "should_trigger": true },
    { "query": "should we do a catalog shopping ad or just statics?", "should_trigger": true },
    { "query": "Does a document ad make sense for our whitepaper promotion?", "should_trigger": true },
    { "query": "The team can only make two creatives a month - which placements can we actually sustain?", "should_trigger": true },
    { "query": "Thought leader ads for booking demos - will that work?", "should_trigger": true },
    { "query": "Everything we run in the Stories placement dies while feed does fine - format problem?", "should_trigger": true },
    { "query": "We repurposed our YouTube ad for TikTok as-is, is that fine?", "should_trigger": true },
    { "query": "What's the right ad unit for an impulse-buy product under $30?", "should_trigger": true },
    { "query": "Client says attention spans are 8 seconds so we should only run 6 second ads - true?", "should_trigger": true },
    { "query": "Is our funnel stage wrong for carousel ads?", "should_trigger": true },
    { "query": "Format audit before we scale spend next month", "should_trigger": true },
    { "query": "Non-skippable vs skippable for our conversion campaign?", "should_trigger": true },
    { "query": "Can we use collection ads for lead generation?", "should_trigger": true },
    { "query": "Which ad formats require a product feed?", "should_trigger": true },
    { "query": "Our ads got approved but aren't delivering on some placements - creative size issue?", "should_trigger": true },
    { "query": "App install campaign - do regular video ads work or do we need a special format?", "should_trigger": true },
    { "query": "What can we run with just product photos and no video?", "should_trigger": true },
    { "query": "Do we need vertical cuts or can automatic placements handle our 16:9?", "should_trigger": true },
    { "query": "Is using a lead form killing our sales-call bookings?", "should_trigger": true },
    { "query": "Which platform should we advertise on, Google or Meta?", "should_trigger": false },
    { "query": "Write me 10 headline variants for our search ads", "should_trigger": false },
    { "query": "Draft a creative brief for our new designer", "should_trigger": false },
    { "query": "Write a UGC script for a skincare testimonial video", "should_trigger": false },
    { "query": "Score these five video hooks and tell me which to test", "should_trigger": false },
    { "query": "How much budget do I need for a proper creative A/B test?", "should_trigger": false },
    { "query": "Our best ad's CTR has been dropping for two weeks - is it fatigued?", "should_trigger": false },
    { "query": "Check whether our conversion pixel is firing correctly before launch", "should_trigger": false },
    { "query": "Our ads get clicks but the landing page doesn't convert - audit it", "should_trigger": false },
    { "query": "Why did our whole account's CPA double last month?", "should_trigger": false },
    { "query": "Build our audience targeting tiers from this ICP", "should_trigger": false },
    { "query": "Target CPA or max conversions bidding for this campaign?", "should_trigger": false },
    { "query": "Are we on pace to spend our $30K monthly budget?", "should_trigger": false },
    { "query": "How should we split $50K between prospecting and retargeting?", "should_trigger": false },
    { "query": "When can we double the budget on our winning campaign?", "should_trigger": false },
    { "query": "Design a retargeting sequence for cart abandoners", "should_trigger": false },
    { "query": "Which customers should seed our lookalike audience?", "should_trigger": false },
    { "query": "Platform says 200 conversions, CRM says 90 - reconcile this", "should_trigger": false },
    { "query": "Mine our search terms report for negative keywords", "should_trigger": false },
    { "query": "What ads are our competitors running right now?", "should_trigger": false },
    { "query": "Plan a campaign promoting our CEO's LinkedIn posts", "should_trigger": false },
    { "query": "How do we write copy for ads inside ChatGPT?", "should_trigger": false },
    { "query": "Is a $210 CAC healthy for our subscription box?", "should_trigger": false },
    { "query": "Set a maximum allowable CAC policy for the company", "should_trigger": false },
    { "query": "We have 40 campaigns - which should we merge?", "should_trigger": false },
    { "query": "Map the buying committee for our enterprise deal ads", "should_trigger": false },
    { "query": "How do I break into paid media as a career?", "should_trigger": false },
    { "query": "Interview questions for hiring a media buyer", "should_trigger": false },
    { "query": "Which PPC newsletters should I follow?", "should_trigger": false },
    { "query": "What video format should I export my YouTube video in for best quality?", "should_trigger": false },
    { "query": "Convert my ad video from MOV to MP4", "should_trigger": false },
    { "query": "What image file format does the ads manager accept, PNG or JPEG?", "should_trigger": false },
    { "query": "Design a data format for our campaign reporting export", "should_trigger": false },
    { "query": "Which ad network pays best for monetizing my blog?", "should_trigger": false },
    { "query": "Rework the hook in the first 3 seconds of this script", "should_trigger": false },
    { "query": "How long should my retargeting window be?", "should_trigger": false }
  ]
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall