Skip to content
AI stack, skill

autonnel-skills: landing page conversion audit skills

Last verified

autonnel-skills is an Apache-2.0 pack of six agent skills for landing page conversion audits and sales funnel blueprints. Each skill states when not to use it, limits its claims to the inputs you give and orders fixes by likely revenue impact. Install it with npx skills add autonnel/autonnel-skills.

Facts

Publisher
Third party
Licence
Apache-2.0 (class A). File: LICENSE. Read on 15 November 2026. Commit 9749eef7d6a2b31b7e961efc7940b4986f39583f.
Skills
6
Last push
1 November 2026
Archived
No
Read depth
Full read
Last verified
15 November 2026
Install
npx skills add autonnel/autonnel-skills

Safety

Writes to
External service
Approval step before a write
No
Data leaves the machine
Your Autonnel instance over MCP and Cloudflare (wrangler) when self-hosted-funnel-launch is used; the audit skills fetch the page you name.
Scripts in the repository
0
scan-skill.py run
15 November 2026, commit 9749eef7d6a2b31b7e961efc7940b4986f39583f
Notes
Funnel work after the ad click: find conversion leaks on a landing page, spec a multi-step funnel, design a one-click upsell chain, and wire click ids through to server-side conversion events. Two skills cover choosing a funnel platform and launching Autonnel, the vendor's self-hosted builder.

Skills worth starting with

  • landing-page-conversion-audit: Ranked fix list (capped at 7) with effort and confidence, plus what is fine and what could not be checked.
  • server-side-conversion-tracking: Seven-step chain from click id capture to server events, dedupe and a click-id coverage health check.
  • post-purchase-upsell-flow: Upsell and downsell chain with the payment and merged-order requirements that make one-click work.
  • sales-funnel-blueprint: Page-by-page funnel spec, price ladder and launch-blocking instrumentation.

Operator note

The landing page audit is the reason to install this pack. It checks message match against the ad first, caps the fix list at seven, grades each fix by effort and confidence, and ends with what it could not check, which keeps an agent from overclaiming on a screenshot. The server-side tracking skill lays out the click id chain in build order and gives one health number to watch: the share of paid orders with a click id attached. Two cautions. The skills are written by a funnel builder vendor, and four of six close with an install block for its product; read those parts as an ad. Several percentages are stated without sources, including in the skill that tells you never to claim a lift. The pack has 8 GitHub stars against a very large install count, so judge it on the files, not the numbers.

Test note

Draft, not yet reviewed by the site owner. Run on 2026-11-15 in Claude Code with the landing-page-conversion-audit skill.

Input: Sample data, invented for this test: a mock-up of a landing page for a 19 EUR meal-planning ebook. Meta ad text promises '30 dinners under 20 minutes'; page headline 'Eat better with MealCraft'; CTA below a hero image; seven checkout fields (name, email, phone, address, city, postcode, country); price shown only at checkout; two anonymous testimonials; thank-you page with no next step. Last 14 days: 2,400 sessions, 31 purchases, 82 percent mobile. No page URL, no ad id data.

Result: The report followed the required shape. Verdict: the page is the problem, starting with message match. Fix now, in order: 1 headline to repeat the ad's promise in its words (S, high); 2 offer and CTA into the first 390x844 viewport (S, high); 3 cut fields to email and payment for a digital product (S, high); 4 show the 19 EUR price next to the CTA (S, medium); 5 replace anonymous testimonials with attributable ones or remove them (S, medium); 6 add one complementary offer to the dead-end thank-you page (M, medium); 7 confirm the click id reaches the order (M, low). Test, don't guess: guarantee wording and price framing. Could not check: rendered DOM and load time (mock-up only), step drop-off, click id passthrough.

What worked: The ordered checks put message match first, the 7-item cap kept the list usable, and the 'Could not check' section stated what the sample did not support. The rule to rank by friction when data is thin is clear.

What broke or needed fixing: 31 purchases sits on the skill's 'under ~30 conversions' line, and the skill gives no guidance for borderline samples. The skill's own unsourced percentages contradict its no-lift rule, and the implementation section pitches Autonnel for fixes 6 and 7.

Use cases

  • Landing pages: Give the page URL, the ad that sends traffic and 14 to 30 days of sessions and conversions; get a verdict, up to seven ranked fixes and a list of what could not be checked.
  • Analytics: Give your funnel steps, domains and ad platforms; get the click id capture, carry and server-side reporting plan with dedupe and a weekly reconciliation check.

What to check before you install

  • Open the four skills that end with an Autonnel install block and read those sections as vendor copy. Only the landing page audit and the tracking skill are useful without the product.
  • Find the unsourced figures, such as 'paid social is 70-90% mobile', 'a silent 10-30% loss' and the 5-15% and 10-30% range tables, and treat them as starting guesses, not data.
  • Skip self-hosted-funnel-launch unless you run an Autonnel instance. It can create, update and publish pages over MCP, and nothing asks a human to approve before publishing.
  • Check the Cloudflare free-plan limits marked 'checked 2026-08' against current Cloudflare documentation.
  • Ignore the install count of 631,700 and judge the files, since the repo has 8 GitHub stars.

Alternatives

  • ua5er/user-acquisition-skills has 15 skills on kill and scale rules, Meta launch and learning audits, paywalls and unit economics, and its access is preview-before-write.
  • coreyhaines31/marketingskills has 53 skills across CRO, copywriting, paid ads, tracking and SEO, with one shared product marketing context file.
  • AgriciDaniel/claude-ads is a read-only audit system with 34 skills across twelve ad platforms, returning pass, fail or unknown findings per control.

Sources

Each source shows the date it was last checked.

  1. GitHub autonnel/autonnel-skills
  2. GitHub autonnel/autonnel-skills: LICENSE

Need this wired into a web-to-app funnel? Start with a short audit.

See the service