Skip to content

mbfinotti/advertising-skills / skills/paid-media-scaling

paid-media-scaling

Verified

Decide when a proven paid campaign has earned a budget increase, how large each step should be, how the ramp sequences over weeks and months, and how to avoid performance collapse on the way up - including readiness gates, vertical vs horizontal scaling, and rollback triggers. Use whenever the user asks whether to scale a campaign, how fast ad spend can rise, mentions a budget ramp, a scaling ceiling, or says performance collapsed after the budget was raised - even if they never say 'scaling'. Covers B2B and B2C. Do NOT use to split a fixed total across campaigns (mbfinotti/advertising-skills@ad-spend-allocation) or to track daily spend against a set budget (mbfinotti/advertising-skills@ad-budget-pacing).

FileKindSize
SKILL.mdskill md29.8 KB
evals/evals.jsonconfig34.7 KB
references/step-size-figures.mdmarkdown10.7 KB
references/worked-ramp-examples.mdmarkdown7.5 KB

evals/evals.json

View on GitHub
{
  "skill_name": "paid-media-scaling",
  "evals": [
    {
      "id": 1,
      "prompt": "I run Meta ads for Velora Skincare. Our prospecting campaign has spent $500/day for the past 8 weeks with a stable $38 CPA against our $55 max. I read that Facebook's official rule is that you can raise budgets 20% every 72 hours without resetting learning, so I want to follow the official rule up to $2,000/day for the holiday push. For context I have the full change history: we bumped this campaign +30% in January, +15% in April, and +25% in June, and I can pull how CPA responded to each. Lay out the schedule.",
      "expected_output": "A ramp response that corrects the '20% every 72 hours' premise as folklore rather than a platform rule, derives the step size from the account's own three past budget changes, sets learning-plus-lag holds, runs readiness checks, and pre-commits a rollback before any schedule.",
      "files": [],
      "expectations": [
        "States that Meta has never published a budget-change percentage and that '20% every 72 hours' is not an official platform rule.",
        "Labels the 20%-every-72-hours figure as folklore or unverified practitioner lore, not documented platform guidance.",
        "Attributes the 20%/72h claim to second-hand practitioner relay of unverified account-rep advice rather than to Meta documentation.",
        "Distinguishes the real platform artifacts behind the folklore (Meta's 'Increase budget by 20%' automated-rule UI preset, and/or Google's 20% guidance applying to bids on Display campaigns only) from an actual budget rule.",
        "States what Meta actually documents: significant edits reset the learning phase, with no budget percentage named.",
        "Derives the recommended step size from the account's own change history (the January +30%, April +15%, June +25% responses) rather than defaulting to a universal percentage.",
        "Sets the hold between steps to at least the learning window plus conversion lag (about a week or more on Meta), not a flat 72 hours.",
        "Runs or requests readiness checks (marginal economics, creative supply, cash, measurement health) before committing to the step schedule.",
        "Pre-commits a rollback trigger, rollback action, and verification date before or alongside the first up-step.",
        "No single step in the proposed schedule raises budget 30% or more in one move without an explicit learning-reset warning.",
        "Any cited step-size or cadence number carries an evidence label (documented, research, or folklore).",
        "If a concrete default percentage such as 15-20% appears, it travels with an explicit instruction to recalibrate from this account's own response data.",
        "The schedule reaches $2,000/day through multiple held steps, and the response names what is monitored during each hold."
      ]
    },
    {
      "id": 2,
      "prompt": "Halberd Outfitters here, DTC menswear. Owner-approved break-even MER is 2.0. Blended MER is currently 2.7 and has been above break-even all summer, so we feel very safe. Main prospecting campaign monthly figures: June $40K spend, $122K revenue. July $48K spend, $140K revenue. August $56K spend, $152K revenue. For September we want to jump straight to $80K since blended has so much cushion. Sanity-check the plan and give us the go-ahead.",
      "expected_output": "A refusal of the go-ahead grounded in marginal-band arithmetic: the newest band returns about 1.5 against a 2.0 break-even even though blended sits at 2.7, so the marginal gate fails and the response stops the ramp and names the fix instead of shipping a smaller one.",
      "files": [],
      "expectations": [
        "Computes marginal return per spend band as delta revenue divided by delta spend, rather than judging the campaign on blended MER alone.",
        "Identifies the June-to-July marginal band as approximately 2.25 ($18K additional revenue on $8K additional spend).",
        "Identifies the July-to-August marginal band as approximately 1.5 ($12K additional revenue on $8K additional spend).",
        "States that the most recent marginal band (about 1.5) is below the owner-approved 2.0 break-even boundary.",
        "Declines to give the go-ahead on the $80K jump despite blended MER of about 2.7 sitting above break-even.",
        "Explains that blended metrics always trail marginal: marginal efficiency turns unprofitable before blended does.",
        "Frames the decision as whether the next dollar of spend still makes money, not how much the account can spend.",
        "Notes the declining trend across bands (about 2.25 down to about 1.5) as evidence efficiency is deteriorating as spend grows.",
        "Treats the failed marginal gate as stop-and-fix: names the failed gate and its fix rather than proposing a smaller budget increase as a consolation.",
        "Flags the proposed $56K-to-$80K move (over 40% in one step) as an oversized single step that risks a learning-phase reset.",
        "Asks whether the revenue figures are platform-attributed, triangulated with business data, or causally measured.",
        "Recommends diagnosing why marginal efficiency is falling (for example saturation signals: frequency, CPM, unique reach, penetration) before adding any budget.",
        "Any stop or exit rule offered references marginal contribution or the marginal band crossing the boundary, not blended MER falling below 2.0."
      ]
    },
    {
      "id": 3,
      "prompt": "Growth lead at Finchley Software, B2B SaaS with self-serve plus sales-assist. Our in-platform numbers: branded search runs an 11x ROAS, retargeting 7.8x, cold prospecting only 2.1x. Everything is platform reporting, we have never run any lift testing. The CFO just approved tripling total paid from $60K to $180K/month, and my plan is to put most of the increase behind branded search and retargeting since those are the proven winners. Draft the scale-up plan.",
      "expected_output": "A plan that refuses to pour the tripled budget into branded search and retargeting on attributed numbers alone: it flags those two lines as exactly where attribution overstates most, requires a causal measurement step for a 3x ramp, and promotes measure-first, without ever claiming the current spend is wasted.",
      "files": [],
      "expectations": [
        "Flags that platform-attributed ROAS overstates incremental value most on retargeting and branded search, the two lines the user wants to scale.",
        "States that a ramp to roughly 3x current spend on purely attributed evidence requires a causal measurement step (holdout or incrementality test) or an explicit user-acknowledged risk line.",
        "Recommends a geo-holdout or incrementality test on branded search and/or retargeting before the bulk of the increase lands there.",
        "References the eBay paid-search experiment (attributed returns in the thousands of percent versus measured negative ROI) as the research-grade warning case for scaling on attribution alone.",
        "Does not claim the account's current branded or retargeting spend is wasted and assigns no percentage of waste: the evidence motivates a test, never substitutes for one.",
        "Classifies the account's current evidence bar as attributed (the weakest tier) and names the upgrade path toward triangulated or causal.",
        "Questions branded search's headroom on demand-capture grounds: search captures existing demand and cannot expand the number of people searching.",
        "Checks retargeting expansion against its finite pool: audience size and frequency limits rather than treating it as freely scalable.",
        "Presents approach candidates and promotes measure-first ahead of immediately laddering budgets, given the 3x target and attributed-only evidence.",
        "Says which condition promoted measure-first (target beyond 2x and/or attributed-only evidence) instead of presenting the choice unexplained.",
        "Notes practitioner geo-holdout findings that reported cold ROAS around 3x often measures nearer 1.8-2.2x incrementally, labeled as vendor or practitioner data.",
        "Any interim budget steps carry pre-committed hold periods and rollback triggers.",
        "Labels cited figures by evidence class: platform numbers as attributed, the eBay result as research, holdout benchmarks as vendor or practitioner data.",
        "The plan does not concentrate the entire increase on branded search and retargeting: some of the allocation logic accounts for the demand-capture ceiling or directs growth toward demand creation."
      ]
    },
    {
      "id": 4,
      "prompt": "We sell compliance software - ACV $28K, roughly 4 months from first click to closed deal. Six weeks ago we scaled LinkedIn from $12K to $20K/month and it's working great: CPL dropped from $95 to $61. I want to push to $45K/month next month, and we'll evaluate each budget bump after one week so we can move fast. Two side notes: sales mentioned discovery calls have felt lighter lately, but they always complain. And LinkedIn reports more conversions than our CRM shows - I trust LinkedIn's numbers since the CRM always lags. Build the ramp.",
      "expected_output": "A response that treats the falling CPL as a possibly broken proxy and the sales remark as a lead-quality decay signal, judges steps on cost per SQL and lead quality over month-plus holds instead of one week, makes the CRM the arbiter over the platform, and gates the push to $45K on verifying quality first.",
      "files": [],
      "expectations": [
        "Flags the falling CPL ($95 to $61 after scaling) as a possible broken proxy to verify, not as proof the scale-up is working.",
        "Treats the sales team's 'discovery calls feel lighter' remark as a lead-quality decay signal to investigate rather than dismissing it.",
        "Requires steps to be judged on cost per SQL and/or a lead-quality score, never on CPL alone.",
        "Rejects the one-week evaluation window: holds must cover learning plus conversion lag, and B2B steps are judged on windows of a month or more using leading indicators.",
        "Reverses the user's stated preference: when platform-reported conversions and the CRM disagree, the CRM wins.",
        "Requires the offline conversion loop (CRM stage changes fed back to the platform) closed before scaling further on any lead metric.",
        "Recommends reconciling platform conversions against the CRM on a recurring monthly basis.",
        "Freezes or gates the push to $45K on verifying lead quality first: the proxy check precedes the raise.",
        "Flags the $20K-to-$45K move (more than doubling in one step) as oversized and replaces it with multiple held steps.",
        "Checks whether sales can absorb the added lead volume without speed-to-lead degrading.",
        "Checks cash absorption: ad spend bills in days while B2B pipeline revenue lags by months (up to the 60-281 day range), so working capital must cover the gap.",
        "Defines a per-step monitor set built on leading indicators (cost per SQL, lead-quality score) with closed-won as the lagging verdict.",
        "Pre-commits a lead-quality-keyed rollback, for example cost per qualified lead above 1.5x target triggering a 20-30% cut, stabilization, then a slower resume.",
        "Anchors affordability in a per-cohort or per-deal payback calculation from the $28K ACV and margin, or asks for the owner-approved CAC boundary, rather than inventing a target."
      ]
    },
    {
      "id": 5,
      "prompt": "The board committed us to $300K/month of paid media by end of Q3 - we're at $95K today. Ecommerce, Meta plus Google, blended ROAS 3.1, all measurement is platform reporting plus GA4. 30-day penetration on our core audience is about 12%, and our in-house studio ships 10 new creatives a month. My plan is simple: raise budgets 20% every week on both accounts until we hit the number. Write that up as a formal plan I can hand to the board.",
      "expected_output": "Instead of formatting the user's ladder, the response interviews, brainstorms the three approaches with an explicit ranking, promotes measure-first because the 3.2x target rides on attributed evidence, and delivers a structured ramp-plan artifact with gates, dated held steps, rollbacks, ceilings, and an exit, pending the user's approval.",
      "files": [],
      "expectations": [
        "Does not simply format the user's 20%-weekly ladder into a plan: the approach choice is examined before drafting.",
        "Asks the user clarifying questions before finalizing the plan, covering at least one of: hardness of the Q3 date, one-off target versus compounding asset, or the effort ceiling.",
        "Presents vertical ladder, measure-first, and horizontal expansion as named candidate approaches with trade-offs, and requests the user's pick.",
        "States the candidate ordering explicitly rather than leaving the ranking implied by list position.",
        "Promotes measure-first given the target is over 2x current spend (about 3.2x) on platform-attributed evidence, and names which condition triggered the promotion.",
        "Re-ranks using account specifics: the in-house studio shipping 10 creatives a month is cited as lowering horizontal expansion's effort.",
        "Applies the penetration bands: about 12% is under the 25% line, so vertical headroom remains available.",
        "The ramp includes a causal measurement step (geo holdout or incrementality test) or an explicit user-acknowledged risk line for scaling to 3x on attributed evidence.",
        "Each proposed step carries a date or hold-until point, a monitor set, a rollback trigger, and a rollback action.",
        "The flat raise-and-judge-weekly cadence is not endorsed as-is: hold length is tied to learning window plus conversion lag, with longer evaluation windows noted where Google (for example Performance Max) is involved.",
        "The plan names its exit condition and which ceiling is expected to bind first.",
        "The deliverable follows a structured ramp-plan shape containing at least gates, evidence bar, approach, steps, ceilings, and exit or open items.",
        "The plan is presented for the user's section-by-section validation or explicit approval, and the assistant does not execute or claim to execute budget changes.",
        "Numbers carry documented/research/folklore labels, and any default step percentage travels with a recalibrate-from-account-history instruction."
      ]
    },
    {
      "id": 6,
      "prompt": "Bit of an emergency. We doubled our Google Ads budget on the 3rd - $800 to $1,600/day - on our roofing-leads campaign. It's the 8th and CPA has doubled from $90 to $180, although that's on just 7 conversions so far. Click to booked job normally takes about 3 weeks for us. My plan: kill the campaign today, then starting tomorrow make one fix per day - Monday tighten locations, Tuesday new headlines, Wednesday adjust the bid target - so we can isolate what actually works. Sound right?",
      "expected_output": "A response that calls the doubled CPA on 7 conversions with a 3-week lag noise, refuses the same-day kill and the one-fix-per-day sequence, reverts to the prior budget as the smallest reversible action, batches any further edits into one session, and pre-commits rollback rules before a slower re-approach.",
      "files": [],
      "expectations": [
        "States that a doubled CPA on 7 conversions is too small a sample to conclude performance collapsed.",
        "Notes the roughly 3-week conversion lag means conversions attributable to the new spend have not landed yet, so the 5-day CPA read is structurally overstated.",
        "Checks or asks about other false-positive causes before acting: tracking outages, seasonality, or running experiments.",
        "Declines to kill the campaign today, reserving instant cuts for genuine emergencies (runaway spend, policy or legal exposure, broken destination, confirmed tracking corruption) that this situation does not match.",
        "Recommends the smallest reversible action, such as reverting to the prior $800/day budget, over pausing or killing the campaign.",
        "Rejects the one-fix-per-day sequence: drip-feeding edits across days can trigger repeated learning resets, and changes should be batched into one session.",
        "Corrects the isolation logic: sequential daily edits during learning-phase instability do not isolate causes.",
        "Identifies the original single +100% jump as an oversized step likely to have reset learning and contributed to the instability.",
        "The re-approach plan uses smaller held steps with a slower resume (on the order of +10% per period after stabilization) rather than re-doubling.",
        "A rollback trigger, rollback action, and verification date are pre-committed before any renewed up-move.",
        "Any judgment window for the re-approach covers the learning window plus the roughly 3-week lag, not another 5-day read.",
        "Distinguishes the persistence-based rollback convention (an efficiency drop persisting about 5-7 days at adequate sample before reverting) from an instant same-day kill, labeling it practitioner convention rather than platform rule.",
        "Monitoring during the re-approach includes leading indicators and delivery/learning status rather than CPA alone inside the lag window."
      ]
    },
    {
      "id": 7,
      "prompt": "DTC supplements brand. One hero Meta prospecting campaign at $70K/month that we've scaled all year, still hitting our 2.2 target MER on blended. Latest 30-day data: audience penetration 37%, frequency 3.6 and climbing, CPMs up 31% year over year, CTR flat, and unique reach has now declined two months in a row. I want to take this same campaign to $140K over the next 8 weeks with 15% weekly bumps. We hold 9 proven ads and can produce 6 new ones a month. Lay out the ramp.",
      "expected_output": "A refusal of the vertical ramp: penetration 37% with declining unique reach, elevated frequency, and rising CPM at flat CTR puts the audience past the 35% band, so vertical is deleted with its constraint named and the increase is redirected to horizontal expansion at proven per-unit budgets with its own creative and overlap checks.",
      "files": [],
      "expectations": [
        "Reads 37% penetration against the saturation bands (under 25% headroom remains, 25-35% hold, 35%+ scale horizontally) and refuses the requested vertical ramp on this campaign.",
        "Treats two consecutive months of declining unique reach as the leading saturation indicator.",
        "Notes frequency 3.6 sits above the roughly 3.0 prospecting danger band, labeling the band as practitioner folklore rather than platform rule.",
        "Reads rising CPM with flat CTR as saturation corroboration rather than as a bidding or creative problem.",
        "States that doubling budget grows penetration only about 50-70%, not 100%, so the target cannot be bought vertically on this audience.",
        "Deletes the vertical ladder for this account rather than merely demoting it, and names the saturation constraint that deleted it.",
        "Redirects the increase to horizontal expansion: new audiences, geos, placements, or channels at proven per-unit budgets.",
        "Notes each new horizontal unit runs its own learning phase and permanently multiplies creative demand, and checks the 6-new-ads-per-month pipeline against that demand.",
        "Warns about audience overlap or cannibalization between new and existing units and calls for exclusions.",
        "Grounds horizontal expansion in penetration-and-light-buyer growth evidence, citing Ehrenberg-Bass or labeling the rationale as research-supported.",
        "Does not treat the campaign still hitting 2.2 blended MER as license to push vertically: asks for or computes marginal efficiency on the newest spend bands.",
        "Each horizontal unit or step still carries a hold period, monitor set, and rollback trigger.",
        "Names an exit condition for the ramp (such as marginal contribution crossing zero or a penetration band) rather than running increases indefinitely."
      ]
    },
    {
      "id": 8,
      "prompt": "I just took over paid media at Brontide Labs - B2B, webinar-lead campaigns, mostly LinkedIn and Meta. Spend is $25K/month, and I'm expected to be at $75K/month within three months; the CFO already signed off on the media budget. We have 4 ads that reliably work, and a freelancer produces about 2 new ones a month. We pay the ad accounts by card weekly, customers pay net-60. One quirk: the account has never had a budget change - same daily budgets since launch in March. Marginal numbers look decent and our tracking audit scored 12 out of 15. Give me the step plan.",
      "expected_output": "A gate-first response: creative supply fails (about 15 proven ads needed at $75K against 4 on hand, and a 2-per-month pipeline cannot close it in time), so the ramp stops and names the fix; cash absorption is flagged on the weekly-card versus net-60 mismatch; and with an empty change log the step size starts at the labeled 15-20% folklore default with a recalibration instruction.",
      "files": [],
      "expectations": [
        "Runs the creative-supply gate arithmetic: roughly 15 proven ads needed at $75K/month (monthly budget divided by $5,000) against the 4 on hand.",
        "Labels the budget-divided-by-$5,000 proven-ad ratio as folklore with an instruction to calibrate it to this account.",
        "Shows the pipeline cannot close the creative gap on the three-month timeline: about 2 new ads per month against a deficit of roughly 11 proven ads.",
        "Treats the failed creative gate as stop-and-fix (fix the creative pipeline, let creative supply set ramp speed) rather than silently shipping a smaller ramp.",
        "Flags the payment-terms mismatch (card billed weekly, customers pay net-60) as a business-absorption check, distinguishing CFO media sign-off from approval of the working capital the ramp consumes.",
        "States the absorption principle that a profitable account can still break the company on cash, or quantifies the float the ramp consumes at the target spend.",
        "Because the change log is empty, starts the step size from the concrete default (15-20% per step, never 30% or more in one move), explicitly because there is no account history to read.",
        "Ships the default step with the instruction to recalibrate from the account's own response after the first few steps, moving to history-derived sizing.",
        "Sets holds covering learning plus B2B conversion lag (weeks to a month or more), judged on leading indicators rather than closed revenue.",
        "Every step carries a pre-committed rollback trigger, rollback action, and verification date.",
        "Step-size and cadence figures carry documented/research/folklore labels, with nothing presented as a platform rule.",
        "Steps are judged on lead quality or cost per qualified lead rather than raw cost-per-lead alone.",
        "Names the exit condition and identifies creative supply as the first-binding ceiling given the arithmetic."
      ]
    }
  ],
  "trigger_queries": [
    { "query": "Our Meta prospecting campaign has held a $42 CPA for two months at $300/day. How fast can we get it to $1,500/day?", "should_trigger": true },
    { "query": "Can we double our Google Ads budget without wrecking performance?", "should_trigger": true },
    { "query": "Build a ramp plan to take paid social from $20K to $80K a month.", "should_trigger": true },
    { "query": "The board wants ad spend at $250K/month by Q2 and we're at $70K. Plan the ramp.", "should_trigger": true },
    { "query": "We tripled the budget last week and ROAS fell off a cliff. How do we re-approach?", "should_trigger": true },
    { "query": "Is it safe to increase the budget on our winning campaign yet?", "should_trigger": true },
    { "query": "How much can I raise my TikTok ad budget each week without resetting learning?", "should_trigger": true },
    { "query": "When has a campaign earned a bigger budget?", "should_trigger": true },
    { "query": "Is the 20% every 72 hours rule for Facebook budgets real?", "should_trigger": true },
    { "query": "Performance collapsed right after we raised spend. What went wrong and what now?", "should_trigger": true },
    { "query": "We keep hitting a ceiling around $50K/month on paid - every push above it tanks efficiency.", "should_trigger": true },
    { "query": "wanna pour way more money into this ad campaign, how fast can i go", "should_trigger": true },
    { "query": "What's a safe weekly budget increase for a LinkedIn campaign that's beating its CPL target?", "should_trigger": true },
    { "query": "Our CAC holds at target - should we scale vertically on the same campaign or expand to new audiences?", "should_trigger": true },
    { "query": "Draft a 12-week budget ramp for our lead-gen campaign, $8K to $30K a month.", "should_trigger": true },
    { "query": "My agency says raise budgets 15% every 4 days. Should I trust that cadence?", "should_trigger": true },
    { "query": "CEO wants us to 4x the ad budget by Black Friday. Talk me through doing that safely.", "should_trigger": true },
    { "query": "The campaign is printing money - how hard can we push it before it breaks?", "should_trigger": true },
    { "query": "At what point do I stop increasing this campaign's budget?", "should_trigger": true },
    { "query": "How long should I wait between budget increases on Meta?", "should_trigger": true },
    { "query": "We raised the daily budget and the CPA doubled. Should we roll back or wait it out?", "should_trigger": true },
    { "query": "What are the signs a campaign is ready for more budget?", "should_trigger": true },
    { "query": "Scale plan needed: $5K/day to $12K/day on our search campaigns before the seasonal peak.", "should_trigger": true },
    { "query": "If I bump the budget 50% in one go, will the algorithm freak out?", "should_trigger": true },
    { "query": "Our investor wants to know how quickly the paid channel can absorb another $1M a year.", "should_trigger": true },
    { "query": "Every time we scale past $2K/day the results fall apart. Why, and how do we break through?", "should_trigger": true },
    { "query": "Give me a step schedule and rollback rules to grow our top campaign's budget 3x.", "should_trigger": true },
    { "query": "Do bigger budgets always mean worse efficiency, or can we grow spend without CPA creeping up?", "should_trigger": true },
    { "query": "We just got extra budget approved for the quarter - how fast can the winning campaigns take it on?", "should_trigger": true },
    { "query": "How do I know if my audience is saturated before pushing more spend into the campaign?", "should_trigger": true },
    { "query": "should i be scaling this adset slowly or just double it, it's been profitable 6 weeks", "should_trigger": true },
    { "query": "We're spending $30K/month profitably. What has to be true before we go to $100K?", "should_trigger": true },
    { "query": "Increase ad spend without killing ROAS - what's the playbook?", "should_trigger": true },
    { "query": "Our Shopify store's ads do 3.5x. I want to go from $1K/day to $5K/day this month - realistic?", "should_trigger": true },
    { "query": "What monitoring should be in place while we ramp campaign budgets up?", "should_trigger": true },
    { "query": "When we grow the budget, should the extra money go on the proven campaign or into new geos?", "should_trigger": true },
    { "query": "Client asked for an aggressive spend ramp next quarter. What guardrails do I build into the ramp itself?", "should_trigger": true },
    { "query": "Facebook rep told us we can raise budgets 20% a day. Is that legit?", "should_trigger": true },
    { "query": "How do we scale our best campaign without triggering the learning phase reset?", "should_trigger": true },
    { "query": "Need to get from 50 leads a month to 200 through paid - how quickly can spend ramp to support that?", "should_trigger": true },
    { "query": "After doubling spend, do we judge results after a week or wait longer?", "should_trigger": true },
    { "query": "Plan the budget increases for Q4 so we peak at $600K/month by November.", "should_trigger": true },
    { "query": "Marginal ROAS vs blended - which one tells me if I should keep raising the budget?", "should_trigger": true },
    { "query": "we've been stuck at the same ad budget for a year, boss finally said grow it, where do i start", "should_trigger": true },
    { "query": "What's the biggest budget jump you can make in one move without blowing up delivery?", "should_trigger": true },
    { "query": "Should I pause the scale-up now that CPMs are rising and reach is flat?", "should_trigger": true },
    { "query": "Our board deck needs a paid-media ramp scenario: current $120K, target $400K in 9 months.", "should_trigger": true },
    { "query": "Ads profitable for 3 straight months - I want a schedule of budget steps with dates and checkpoints.", "should_trigger": true },
    { "query": "How do I scale a campaign that's limited by a small B2B audience?", "should_trigger": true },
    { "query": "The last two times we raised budgets, performance collapsed within days. Design a safer ramp this time.", "should_trigger": true },
    { "query": "Is now the right time to scale, or do we need an incrementality test first?", "should_trigger": true },
    { "query": "Take my $15K/month campaign to $60K - what could go wrong on the way up and how do we prevent it?", "should_trigger": true },
    { "query": "How should I split $50K a month between Google and Meta?", "should_trigger": false },
    { "query": "Which campaigns should get more of our fixed quarterly budget?", "should_trigger": false },
    { "query": "Reallocate spend between prospecting and retargeting for better returns.", "should_trigger": false },
    { "query": "Are we on pace to spend the $30K budget this month?", "should_trigger": false },
    { "query": "Campaign is underspending its daily budget - project month-end spend for me.", "should_trigger": false },
    { "query": "Build a tracker that flags when spend runs ahead of budget.", "should_trigger": false },
    { "query": "What's the maximum CAC we can afford given our margins?", "should_trigger": false },
    { "query": "Set a ROAS floor and kill-switch policy for our paid program.", "should_trigger": false },
    { "query": "Who should have authority to approve ad budget changes at our company?", "should_trigger": false },
    { "query": "Is a 2.5 ROAS good for an ecommerce store our size?", "should_trigger": false },
    { "query": "Is our $180 CAC healthy compared to industry benchmarks?", "should_trigger": false },
    { "query": "Should we switch from manual CPC to target ROAS bidding?", "should_trigger": false },
    { "query": "Delivery collapsed after we lowered the target CPA - what's wrong with our bidding?", "should_trigger": false },
    { "query": "Why did our CPA go up? These ads used to work fine.", "should_trigger": false },
    { "query": "Audit our ad account and tell me why performance is declining.", "should_trigger": false },
    { "query": "Frequency is climbing and CTR is dropping on our hero ad - is the creative worn out?", "should_trigger": false },
    { "query": "When should we refresh creative on a long-running campaign?", "should_trigger": false },
    { "query": "We have 40 ad sets stuck in learning limited - should we consolidate campaigns?", "should_trigger": false },
    { "query": "Plan a migration to merge our fragmented campaigns without resetting learning.", "should_trigger": false },
    { "query": "Where should a B2B startup spend its first $20K of ad budget?", "should_trigger": false },
    { "query": "Is connected TV worth adding to our channel mix?", "should_trigger": false },
    { "query": "Check whether our conversion pixel fires correctly before we launch.", "should_trigger": false },
    { "query": "The platform reports 240 conversions but the CRM shows 170 - reconcile the gap.", "should_trigger": false },
    { "query": "Why does Meta claim more revenue than our order system records?", "should_trigger": false },
    { "query": "Ads get clicks but the landing page doesn't convert - audit it.", "should_trigger": false },
    { "query": "Design our cart-abandonment retargeting sequence with frequency caps.", "should_trigger": false },
    { "query": "How long should the retargeting window be for a 30-day sales cycle?", "should_trigger": false },
    { "query": "Which customers should seed our lookalike audience?", "should_trigger": false },
    { "query": "Map our ICP to a layered audience targeting plan.", "should_trigger": false },
    { "query": "Design an A/B test to compare two ad concepts with proper sample size.", "should_trigger": false },
    { "query": "How much budget does a creative split test need to reach significance?", "should_trigger": false },
    { "query": "Mine the search terms report and build our negative keyword list.", "should_trigger": false },
    { "query": "Write five headline variations for our spring sale ads.", "should_trigger": false },
    { "query": "Score these three video hooks and tell me which deserves test budget.", "should_trigger": false },
    { "query": "Put together a creative brief for our new UGC campaign.", "should_trigger": false },
    { "query": "What ads are our competitors running right now?", "should_trigger": false },
    { "query": "Which ad formats work for mid-funnel B2B on LinkedIn?", "should_trigger": false },
    { "query": "How do I break into paid media as a career?", "should_trigger": false },
    { "query": "Write interview questions for hiring a media buyer.", "should_trigger": false },
    { "query": "Which PPC newsletters and podcasts should I follow to stay current?", "should_trigger": false },
    { "query": "Promote our CEO's LinkedIn posts as thought leader ads.", "should_trigger": false },
    { "query": "Map the buying committee for our enterprise deal and who to target.", "should_trigger": false },
    { "query": "Sponsored answers inside AI assistants - how should our ad copy adapt?", "should_trigger": false },
    { "query": "Scale our content production from 4 to 20 blog posts a month.", "should_trigger": false },
    { "query": "We need to scale our outbound sales team from 3 to 15 SDRs - plan the ramp.", "should_trigger": false },
    { "query": "How fast can we scale our email send volume without hurting deliverability?", "should_trigger": false },
    { "query": "Scale our influencer program from 5 to 50 creators next quarter.", "should_trigger": false },
    { "query": "How do we scale our affiliate program's payouts as revenue grows?", "should_trigger": false },
    { "query": "Increase our SEO content budget - how should we phase the investment?", "should_trigger": false },
    { "query": "What's a good budget for launching our first ever ad campaign?", "should_trigger": false },
    { "query": "Set the initial daily budget for a brand-new campaign still in learning.", "should_trigger": false },
    { "query": "Spread this quarter's $200K media budget across our five product lines.", "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: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated 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: 323kUpdated Claude CodeCodexVerifiedInstall