Skip to content

mbfinotti/advertising-skills / skills/ad-copy-variants

ad-copy-variants

Verified

Turn 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).

FileKindSize
SKILL.mdskill md23.0 KB
evals/evals.jsonconfig29.7 KB
references/angle-library.mdmarkdown10.1 KB
references/variant-matrix-template.mdmarkdown4.8 KB
references/worked-examples.mdmarkdown5.7 KB

references/angle-library.md

View on GitHub
# Angle Library

An angle is the argument for why this reader should care - a distinct motivation, stated as a human sentence. Changing the angle changes who the ad attracts; changing the wording only changes how one argument sounds.

## Core angle taxonomy

Eight recurring angle families, each a different reason someone clicks. Anchor every candidate angle to evidence the user supplied; an angle without evidence is a guess wearing a label.

Rows run in efficiency order - value per unit of effort, never cheapest-first. The other axes disagree, so each gets its own line:

- **efficiency**: pain > mechanism > outcome > identity > social proof > contrarian > comparison > urgency
- **proof-asset cost**: social proof > outcome > comparison > contrarian > mechanism > urgency > pain == identity
- **compliance cost**: comparison > urgency > contrarian > social proof > outcome > mechanism > pain == identity

Pain and identity tie on both cost axes, and for the same reason on each: both are written straight out of evidence already in hand, so no asset gets acquired and no checkable claim gets made that anyone could ask you to substantiate.

| Angle            | The argument it makes                                                | Proof it costs you                                                                       | Compliance cost                                                                         | Leans B2B / B2C                              |
| ---------------- | -------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------- | -------------------------------------------- |
| Pain point       | "You have this specific problem, and it costs you."                  | Minutes - mined from the interview evidence                                              | None past the grounding rule                                                            | Both - the default anchor                    |
| Mechanism        | "Here is _why_ this works when alternatives didn't."                 | About an hour with whoever knows how the product works                                   | The mechanism must be true as stated                                                    | Both; strongest in sophisticated markets     |
| Outcome          | "Here is the measurable after-state."                                | Waits on a measured before/after somebody has to produce                                 | Substantiation review for every figure and superlative                                  | Both; B2B wants the number quantified        |
| Identity         | "This is what people like you use."                                  | Minutes - the segment's language is already in the evidence                              | None past the grounding rule                                                            | Mostly B2C; B2B as team/craft pride          |
| Social proof     | "People like you already switched, and said this."                   | Days to weeks - testimonials plus usage permission; the one asset you cannot manufacture | Written permission per quote on file                                                    | Both; B2B needs a named role or badge        |
| Contrarian       | "The common advice is wrong; here's the uncomfortable truth."        | About an hour, if the belief is already held and evidenced internally                    | Brand sign-off - the edge is the risk                                                   | Both; requires proof to avoid empty edge     |
| Comparison       | "Against the way you do it now (or a rival), here's the difference." | Every claim about the rival substantiated one by one                                     | Least reversible in the set: a trademark complaint pulls it mid-flight, after the spend | Both; check trademark constraints first      |
| Urgency / stakes | "The cost of waiting is real and near."                              | Near-zero where a real deadline exists; unavailable where it doesn't                     | A manufactured deadline fails platform review                                           | B2C offers; B2B renewal/compliance deadlines |

Value is deliberately not ranked across families. Which family pays best is decided by awareness stage and market sophistication below, not by the family itself - a global value order would be false precision.

### What this order starves

**Social proof, and outcome behind it.** Both sit mid-table on efficiency purely because their assets are expensive - days to weeks for a permission-cleared testimonial, a measured before/after somebody has to produce. The value table above puts them among the strongest arguments in the set, and social proof is the one asset that cannot be manufactured at any price. An account that only ever computes efficiency drafts pain and identity every round and never commissions the proof, so the ceiling on what its ads can argue never moves.

Promote them anyway when any of these holds, and say which one did it:

- The asset already exists - a signed case study, a cleared testimonial, a measured result sitting unused. Its cost is already paid, so it jumps the queue outright.
- The user answered the interview with a compounding mandate. The asset outlives the round that buys it and re-enters every later ranking at near-zero.
- The traffic is product-aware or the market is sophisticated. Proof and comparison are what those readers weigh; a pain angle they have already accepted buys nothing.
- Nothing else is left. Two rounds where pain and identity both flatten means the cheap arguments are exhausted, not that the market is.

### Ruled out is deleted, not demoted

An angle the user's constraints forbid leaves the menu and is named as deleted in the angle bank. Never park it at the bottom of the order, where the next round reads it as merely unpopular and quietly briefs it back in as scope.

- No cleared testimonial and no permission to obtain one → **social proof is deleted**, not last.
- Naming competitors refused, or the trademark exposure unacceptable → **comparison is deleted**.
- No real deadline or scarcity → **urgency is deleted**; a manufactured one fails platform review anyway.
- Brand sign-off unavailable → **contrarian is deleted**; the edge is the whole angle, and an unapproved edge gets sanded off in review.
- A regulated category barring the claim the angle rests on → that angle is deleted for this account, whatever it scores.

Write the deletion and its cause into the angle bank, so a later round re-tests the constraint rather than the angle.

### Other re-ranks

That order is a default, not a law: it shifts with context and with who executes it. Re-rank against this account, and against the constraints answered in the skill's interview, before drafting.

- A hard deadline promotes pain, identity, and urgency; they draft the same day.
- Mechanism moves up on solution- and product-aware traffic; pain moves down at most-aware, where the offer leads.
- A team with an in-house researcher, a public reference customer, or counsel on call pays less than this default assumes for outcome, social proof, and comparison.

## Awareness stages decide the opening

Eugene Schwartz's five stages (Breakthrough Advertising, 1966). The angle stays constant across stages; its expression changes. One stage per variant set unless the user splits traffic by stage.

- **Unaware** - open on the symptom or an arresting truth; naming the product first is wasted.
- **Problem-aware** - open on the problem in the customer's own words (PAS territory); the reader recognises themselves before meeting the product.
- **Solution-aware** - open on the mechanism or differentiator: why this way beats the category's known way.
- **Product-aware** - open on proof and comparison: numbers, named results, objection handling.
- **Most-aware** - open on the offer and the CTA; anything else delays the click.

Companion concept - **market sophistication**: how many times this market has already heard claims like yours. In a fresh market a plain claim works; in a tired one the claim must escalate to mechanism, proof, or identity, because the reader has heard "faster and easier" too many times to register it.

## B2B vs B2C angle skew

- **B2C sets** draw from these, in efficiency order: pain > identity > outcome-as-transformation > social proof > urgency/offer. The reader is the buyer; emotion and a single-decision close are legitimate.
- **B2B sets** draw from these, in efficiency order: quantified outcome > mechanism > risk/compliance reduction > peer proof > total cost of ownership. The reader often builds an internal case for a committee - a strong B2B angle arms that reader with a line they can repeat in a meeting.
- Mechanism leads the B2B order instead whenever no measured number exists yet: quantified outcome is only cheap for an account that has already done the measuring.
- Both orderings are defaults that move with the account - an owned asset outranks a nominally stronger angle whose asset would take a week to obtain.
- This skew is practitioner synthesis, not measured survey data - treat it as a starting bias, and let the user's own evidence override it.

## Angle vs wording - a worked pair

Same value proposition: an invoicing tool that gets freelancers paid faster.

**Three genuine angles** (different problem / mechanism / audience):

1. Pain: "Chasing overdue invoices is a second unpaid job." - the problem is the chasing.
2. Mechanism: "Automatic reminders and card-on-file mean the money moves without you asking twice." - the argument is why it works.
3. Audience/identity: "Built for freelancers who'd rather design than do accounts receivable." - the argument is who it's for.

**Three rewords masquerading as variants** (one angle, three coats of paint):

1. "Get paid faster."
2. "Stop waiting on payments."
3. "Your invoices, paid sooner."

All three make the identical outcome argument; swap any two headlines and nothing breaks. Testing them against each other teaches almost nothing about the market - only about synonym preference.
finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the workStars: 322kUpdated Claude CodeCodexVerifiedInstall
systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixesStars: 322kUpdated Claude CodeCodexVerifiedInstall
test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation codeStars: 322kUpdated Claude CodeCodexVerifiedInstall
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: 198kUpdated Claude CodeCodexVerifiedInstall