mbfinotti/advertising-skills / skills/ad-copy-variants
ad-copy-variants
VerifiedTurn one core value proposition into genuinely distinct ad copy variants - headline, primary text, description, and CTA combinations - each labelled with angle, awareness stage, formula, and version so test results stay attributable. A real variant is a new angle, never a reworded line. Use whenever the user asks to write ad copy, needs headline or CTA variations, wants copy to A/B test, has ads that all sound the same, or needs a search ad headline set or an ML asset pool - even if they never say 'variants'. Covers B2B and B2C. Do NOT use for video or creator scripts (mbfinotti/advertising-skills@ugc-ad-scripts), a designer brief (mbfinotti/advertising-skills@ad-creative-brief), or test design (mbfinotti/advertising-skills@ad-creative-test-plan).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 23.0 KB |
| evals/evals.json | config | 29.7 KB |
| references/angle-library.md | markdown | 10.1 KB |
| references/variant-matrix-template.md | markdown | 4.8 KB |
| references/worked-examples.md | markdown | 5.7 KB |
references/worked-examples.md
View on GitHub# Worked Examples
Two fictional products, one B2C and one B2B. Every proof line in the good examples traces to (fictional) user-supplied evidence, as the grounding rule requires. Limits shown are illustrative - in real work they come from the user's ads manager.
## B2C: language-learning app, social feed
Value proposition: "Conversational in a new language on 15 minutes a day." Traffic: problem-aware. Evidence supplied: app-store reviews. Surface: social feed (headline ~40 chars visible, primary text folds after ~125 chars).
**Variant `pain_pa_01_v1` - angle: pain (the quitting cycle), formula: PAS**
- Headline: `The app you won't quit in week 3`
- Primary text: `Started a language three times, quit three times? The problem isn't you - it's courses built for people with two free hours. 15 minutes a day, built around your commute.` _(fold lands after "quit three times?" - hook complete before it)_
- Description: `15 min a day. Free to start.`
- CTA: `Install now`
- Evidence: reviews R12, R31 - "I always quit around week 3", "finally one I stuck with"
**Variant `identity_pa_02_v1` - angle: identity (the almost-bilingual), formula: BAB**
- Headline: `For the "I studied it in school" crowd`
- Primary text: `You can order a coffee, but the conversation dies after two sentences. Users who did a daily 15 minutes told us the difference: they stopped translating in their head. That's the bridge.`
- Description: `Pick up where school left off.`
- CTA: `Install now`
- Evidence: reviews R7, R19 - "school Spanish", "stopped translating in my head"
**Variant `mechanism_sa_03_v1` - angle: mechanism (spaced speaking practice), formula: 4 Us headline over a proof body**
- Headline: `Speaking practice, not flashcards`
- Primary text: `Most apps drill vocabulary you never say out loud. Ours makes you speak from day one and replays what you got wrong right before you'd forget it. Reviewers call it "the first app that made me talk".`
- Description: `Speak from lesson one.`
- CTA: `Try it free`
- Evidence: review R44 (verbatim quote); mechanism description confirmed by the user
Why the set passes: three different arguments (quitting pain / identity / mechanism), one angle each, every quoted line traceable, key phrases before the fold, all labels complete.
### Negative counter-example - rejected set
- `Learn a language in just 15 minutes a day`
- `Master a new language - 15 minutes daily`
- `15 minutes a day to a new language`
- `Voted the #1 language app` - fabricated: no such award exists in the supplied evidence. Fails proof traceability; the line is a fabricated claim, not a variant.
One outcome claim, three word orders across the first three lines. Swap any two headlines and nothing changes - fails the What Counts test and the one-angle-per-variant check. Also fails the generic test: any competitor with a daily-lesson app could run all three unchanged.
## B2B: expense-management SaaS, professional-network feed
Value proposition: "Close the books five days faster." Buyer: finance manager who must convince a CFO. Traffic: solution-aware. Evidence supplied: one customer case study (mid-market retailer, close time 9 days -> 4), two review-site reviews with permission, a sales-call note ("audit prep is the panic moment"). Surface: professional-network feed - headline clips permanently at ~70 chars, no expander.
**Variant `outcome_sa_01_v1` - angle: quantified outcome, formula: PAS**
- Headline: `Month-end close in 4 days, not 9 - how one retail team did it` _(62 chars - fits the permanent clip)_
- Primary text: `Every extra close day is a day finance can't answer "how did we do?" A mid-market retailer cut close from 9 days to 4 by killing manual receipt-chasing. The case study breaks down what changed, week by week.`
- CTA: `Download`
- Evidence: customer case study (user-supplied, publication approved)
**Variant `risk_sa_02_v1` - angle: risk/compliance, formula: BAB**
- Headline: `Audit season without the shoebox of receipts`
- Primary text: `Right now: three weeks of retroactive receipt-hunting every audit. After: every expense already matched, coded, and export-ready. The bridge is capture at the moment of spend - not at month end.`
- CTA: `Learn more`
- Evidence: sales-call note ("audit prep is the panic moment"); workflow claims confirmed by the user
**Variant `committee_sa_03_v1` - angle: peer proof for the internal case, formula: proof-led**
- Headline: `The one-pager your CFO will ask for anyway`
- Primary text: `"We approved it in one meeting - the ROI sheet did the talking," says a finance manager at a 200-person logistics firm. Get the same ROI template, pre-filled with the 9-to-4-day close data.`
- CTA: `Download`
- Evidence: review RS-A (permission on file); case-study figure as above
Why the set passes: three arguments (outcome / risk / the internal-case reader), each proof point traceable, headlines under the permanent clip, and the third variant is written for what B2B copy actually does - arming the reader to sell internally.
### Negative counter-example - rejected set
- Headline: `Trusted by 10,000 CFOs worldwide` - fabricated: no such figure in the evidence. Fails proof traceability, and would also fail platform substantiation review.
- Headline: `Stop wasting money on expense reports - claim your free audit today before spots run out!!` - manufactured urgency with no real deadline, generic enough for any competitor, and clicks it wins won't convert: the destination page offers a demo, not an "audit". Fails the generic test and message match.
- Headline: `The revolutionary AI-powered future of financial operations excellence and workflow transformation` - 98 chars on a surface that clips permanently at ~70: the argument dies mid-sentence. Says nothing specific; fails surface fit and the generic test at once.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |