Skip to content
AI stack, skill

marketingskills (coreyhaines31): review for paid acquisition

Last verified

marketingskills by Corey Haines is an MIT-licensed pack of 53 agent skills for copywriting, CRO, SEO, analytics, pricing and growth, built around one shared product marketing context file. Install it with npx skills add coreyhaines31/marketingskills. These notes cover which skills fit paid acquisition work and what to check first.

Facts

Publisher
Third party
Licence
MIT (class A). File: LICENSE. Read on 15 November 2026. Commit 1f837447b20bb6f0e72d87b80ece1c221c362c7d.
Skills
53
Last push
11 November 2026
Archived
No
Read depth
Full read
Last verified
15 November 2026
Install
npx skills add coreyhaines31/marketingskills

Safety

Writes to
External service
Approval step before a write
Yes
Data leaves the machine
Only when you wire tools: ad platform, CRM, email and data APIs through tools/clis and tools/integrations with your own keys; reverse-engineer-marketing/scripts/teardown.mjs fetches public pages and crt.sh.
Scripts in the repository
105
scan-skill.py run
15 November 2026, commit 1f837447b20bb6f0e72d87b80ece1c221c362c7d
Notes
A broad marketing library for founders and technical marketers: page CRO, copy, ad creative, paid campaigns, tracking setup, pricing, onboarding, SEO and lifecycle email. One shared product-marketing context file feeds every skill, so the agent asks about your product once.

Skills worth starting with

  • ad-creative: Angles, headline and primary text variations with character counts per platform, plus grounding rules for batch work.
  • ads: Campaign planning, audits and kill or scale calls for Google, Meta and LinkedIn, with draft-first changes on live accounts.
  • conversion-tracking: Sets up or audits conversion tracking and checks that a test conversion arrives; includes GTM import recipes.
  • copywriting: Diagnoses whether input is blank, has a voice or needs a point edit before writing page copy.
  • cro: Page and form conversion diagnosis with quick wins and test ideas.
  • attribution: Explains conflicting attribution numbers and designs first-party measurement.

Operator note

A broad free marketing library that installs in one command; for paid work the ad-creative skill is the part to start with. It asks for context before writing, counts characters per platform, and keeps a list of AI tells that removes the usual generated phrasing. Hypotheses have to name a metric and a control, which makes tests that can lose. Watch three things. The grounding rule against invented proof applies only to batch mode, so on a fresh product the angle table invites social proof you cannot back. Many skills quote benchmark numbers with no source; treat them as folklore until you check them against your own account. Some skills pitch the author's own agent product and paid partners, with disclosure. For app and subscription UA you will still need your own kill and scale rules; this pack is strongest on copy, CRO and tracking setup.

Test note

Draft, not yet reviewed by the site owner. Run on 2026-11-15 in Claude Code with the ad-creative skill.

Input: Sample data, invented for this test: Tidyhome, a fictional subscription app that builds a 10-minute evening cleaning plan, 7-day free trial then 6.99 a month, Meta feed placement, audience of people in small flats who work late. No product-marketing context file, no reviews, no performance data (Mode 1, generate from scratch).

Result: The skill first asked for the context file, then for platform, offer, audience, data and constraints. It produced three angles (pain point, outcome, identity) with three headlines each, all within the 40-character Meta guide (23 to 39 characters, counted by script), and one primary text per angle of 146 to 152 characters with the hook in the opening line. Social proof and urgency angles from its table were dropped because the sample had no reviews and no real deadline. Each angle got a test hypothesis naming CTR against the pain point control, without a magnitude, as the skill asks when no baseline exists.

What worked: The character-limit table and the count check caught nothing over limit and would flag it if they did. The AI-tells list removed two 'It's not X, it's Y' drafts and a stacked negation. The hypothesis rule produced tests that can lose.

What broke or needed fixing: In all three primary texts the trial and price terms landed after the 125 visible characters, because the skill says to front-load the hook and gives no rule for where offer terms go. The no-invented-claims rule sits under batch mode only, so in Mode 1 the angle table pushes toward 'Join 10,000+' style proof the product does not have; the operator has to cut it.

Use cases

  • Ad copy: Give the offer, audience, platform and any winning ads; get angles with headline and primary text variants counted against platform limits, plus a test hypothesis per angle.
  • Landing pages: Give a landing page URL or screenshot and the traffic source; get quick wins, larger changes and test ideas from the cro skill.

What to check before you install

  • Treat benchmark numbers as unverified. Lines such as 'Captions always (85% watch without sound)' and 'AI-referred traffic converts 6-27x higher' carry no source, so compare them with your own account data.
  • Open ad-creative and look at the sample outputs and the angle table. Lines like 'Join 10K+ Teams' and 'save 10+ hours/week' need proof you can back before they go into an ad.
  • Check cold-email by hand. It has no explicit approval line before sending, unlike ads, social and directory-submissions.
  • Read the three skill scripts (teardown.mjs, build_recipe.py, check_recipes.py) and the 26 tracking snippet files before running anything. None were executed in the review.
  • Note the disclosed partners and the Magister pitch, and decide whether that affects what you take from the skills.

Alternatives

  • ua5er/user-acquisition-skills has 15 skills for app and subscription paid UA, with written kill and scale rules and Meta launch audits, where this pack leans toward copy, CRO and tracking setup.
  • AgriciDaniel/claude-ads is a read-only audit system covering twelve ad platforms, returning pass, fail or unknown findings per control, while this pack can create or change campaigns with approval.
  • autonnel/autonnel-skills has six funnel skills for landing page audits, upsells and server-side conversion tracking, and it is tied to the Autonnel vendor for some of its steps.

Sources

Each source shows the date it was last checked.

  1. GitHub coreyhaines31/marketingskills
  2. GitHub coreyhaines31/marketingskills: LICENSE

Want your own AI and acquisition setup checked for gaps and risks? Start with a growth audit.

See the growth audit