Show IH: We built a Claude skill that audits a SaaS landing page in 90 seconds (free, drop your URL)
Hey IH,
We've been reading a lot of indie SaaS landing pages over the last month. The same 12 conversion issues show up almost every time — hero copy promising features instead of outcomes, pricing blocks that hide the real comparison, signup forms asking for company size before email, etc.
So we built a Claude Code skill that runs all 12 checks in roughly 90 seconds. You drop a URL, it returns the 12 issues ranked by impact with a rewrite suggestion for each.
Free version is live on Gumroad while we calibrate it on more pages: https://andersalknes.gumroad.com/l/landing-cro-audit?utm_source=indiehackers&utm_medium=organic&utm_campaign=s019-validate&utm_content=2026-05-07-launch
Two asks:
- Drop your URL in the comments — happy to run it on yours and post the output back. Genuinely useful for us as we lock the framework.
- What 13th issue should be on the list? Curious what indie founders have caught on their own pages that we might be missing.
Built with Claude Code skills. The skill itself is the product — runs locally, no SaaS dashboard, no signup wall.
The 90-second audit is the easy part for most solopreneurs - the hard part is what happens after. Which findings get actioned, in what order, and why you skipped the others.
That's where landing page iteration stalls for most solo founders: not lack of feedback, but no system to decide on and track what you're actually changing. Three audits in, they're back to square one with a different tool.
I've been building a Notion OS for solopreneurs with a Decisions database for exactly this: log what you changed, why, what you expected vs what happened. Turns iteration from gut-feel to something you can learn from.
What insight does the audit surface that founders most reliably act on?
Good question — I've been tracking this. The audits that get re-shared (and the ones where the founder DMs back saying "shipped the fix") almost always come down to one of two patterns:
A trust signal that's missing where buyers look for it. Founder bio buried below the fold, no preview of the actual product, no refund line — the audit surfaces where the friction is, not just that there's friction. Founders act on this fast because it's binary: the proof is either there or it isn't.
A headline that names the deliverable instead of the pain. "X for Y" framings are easy to write but they make the buyer do the translation work. When the audit reframes the headline around the actual moment the pain hits, founders test it within a day.
The audits founders don't act on are the layout / copy-tone ones. Those need taste, and taste is hard to outsource to an AI tool. Useful to know going in — the audit is best as a diagnostic layer for the obvious leaks, not a redesign engine.
Your Decisions database approach sounds like the right shape for the iteration loop after the audit — what triggers a buyer to log a decision vs just close the doc?
The one founders act on almost every time: pricing-tier inversion — where the cheaper tier accidentally looks more capable than the paid one because the comparison table sorts features without sorting outcomes. Easiest to fix (re-order rows + bold one line) and the only issue that moves conversion within a week, so the feedback loop is tight enough that the founder actually believes the audit.
Second one: CTA verb mismatch — hero promises "find your customers," button says "sign up free." That gap costs more than people think, and the rewrite is one word.
Bookmark insight → log rewrite → measure-the-week-after sounds like exactly the loop your Decisions database is built for. Drop a URL of yours if you want to test it — I'll run the 12-check and post back the evidence/guidance split.
Dropping in without the URL since I'm new here — but curious about the "advice vs evidence" framing you landed on.
We've been wrestling with the same split on a workflow guide for vibe coders. "Context management" felt like advice. "Session recovery" feels closer to evidence — something that's either broken or fixed, not a suggestion.
The rename trigger you described ("the audit said this is costing conversions" vs "the audit suggested") is exactly the test I needed. Going to watch whether people say "the system keeps me in context" vs "I never lose context anymore."
What's the 13th issue? I'd guess: hero copy written for the founder, not the frustrated user who just hit the problem for the third time this week.
The split lands like this in practice: "evidence" is anything I can point at on the page and say "this specific element is costing conversions right now." "Consider this" is anything that depends on positioning calls only you can make. Same audit, two confidence levels.
Your reframe is exactly right — "session recovery" is an evidence claim (broken or fixed, observable in any chat) and "context management" is advice (sounds smart, but the buyer can't tell if they have the problem). The first one earns the click.
13th issue, if I had to bet: hero copy written for the founder, not the frustrated user who just hit the problem for the third time this week. Founders write from "what we built." Buyers arrive from "what just broke for me again." The fix is usually a one-paragraph rewrite where the first sentence describes the buyer's last 30 minutes, not the product's category.
Just ran the full 12 on your Vibe OS URL and dropped it in the thread you started below — #2 on that list is exactly this issue.
The 13th-issue guess is sharp — that's exactly the failure mode I keep seeing. The fix is brutal once you name it: re-read the hero cold, pretend you've never built the thing, and check whether it describes a state you'd want or a workflow the founder is proud of. Most hero copy fails that test.
"Context management" mapping cleanly to neither concept nor evidence is the right call. The clean test we're using on the audit: can the issue be written as a binary check a stranger could run? "Pricing tier $X comes before $Y in the table" is evidence — anyone can verify. "Context loss is a problem" is guidance — it lives in interpretation. If your workflow framing fails the binary-check test, it's still useful, but it sits in the second tier.
Happy to run the 12-check on the vibe-coder guide if you've got a URL — output back here, no catch.
Here it is: baykonur.gumroad.com/l/pqkoy
Vibe OS — session recovery system for non-technical vibe coders. $29 one-time. Would genuinely love the 12-check output.
Ran it. Output below — splitting it the way we discussed (evidence = binary, observable; consider this = depends on positioning calls only you can make).
EVIDENCE (high-confidence conversion leaks)
No product visual above the fold — the hero image is the default Gumroad placeholder icon grid. On a $29 digital product, the visual carries the "is this real?" decision. Vibe coders especially decide on screenshots of the actual templates inside. Fix: drop in a real screenshot of the session-protocol template or starter stack inside Claude/Cursor.
The pain hook is buried. "Every time you open a new chat, your AI has no idea what you built yesterday" is the line that earns the click — but it sits as body paragraph 1, below the price strip. Promote it into the subhead so the visitor reads pain → category in one scan.
The "what's inside" list reads as features, not outcomes. "MCP configuration — pre-configured" is a deliverable. Add an outcome clause to each — "…so you stop debugging connector setup every new project."
The "5 attempts" auto-stop script is the most original feature on the page and it's bullet #7. Everyone ships templates and settings; the going-in-circles detector is the differentiator. Pull it up to #1 or feature it visually.
No risk reversal. Low-friction $29 price but no refund line, no preview of any template, no "see the session-end ritual before you buy." Add either a 7-day refund or a screenshot of one real template — the session-end.md ritual is already a link, would be perfect to show.
The founder line is buried. "Rebuilt his project from scratch twice before figuring this out" is the strongest trust signal on the page — it tells the buyer this came from real pain, not theory. Move it directly under the headline as a one-line origin proof.
CONSIDER THIS (lower-confidence — depends on calls you should make)
"Non-technical vibe coders" — narrow on purpose, or accidental? "Vibe coder" already implies non-technical. The double-qualifier might be narrowing TAM more than you mean to. If the system also works for technical solo founders using Claude + Cursor, dropping "non-technical" widens the audience without diluting positioning.
No price anchor. $29 reads cheap, but compared to what? "Cheaper than one month of Cursor, fixes the problem permanently" gives the brain something to compare against.
"Two complete starter stacks — one for mobile, one for micro-SaaS" is potentially the lead bullet for half your buyers. Mobile-app vibe coders are a distinct sub-segment. Worth testing two landing variants or at least a section split where mobile-app builders see the mobile stack first.
The closing "This is for you if:" sentence got truncated on my read of the page. Worth checking it renders fully on mobile — if it cuts off, you're losing the qualification moment right before the CTA.
The session-end.md link is http:// not https://. Minor, but on a trust-sensitive $29 page some browsers will badge it as insecure.
No social proof / count. No "200 vibe coders use this" or "shipped after 3 of my own projects survived the rebuild." If you have any usage number — even your own — surface it. If you don't yet, the founder-rebuilt-twice story IS the proof; lean harder on it.
If I were you I'd ship 1, 2, 4, 6, 11 in that order. The visual + headline restructure is one ~30-min edit and moves the needle more than anything else.
The core issue is the name.
“AI Growth Coach” sounds like a service, not a product.
And for a Claude skill that audits SaaS landing pages, that matters.
You’re not really selling coaching here.
You’re selling fast judgment on conversion leaks.
That should feel sharper, more productized, and more ownable.
If the skill becomes more than a Gumroad file, I’d seriously look at a stronger .com.
Beryxa.com would fit this better.
Short, SaaS-native, and credible enough to carry a conversion/audit product without sounding like another AI consultant.
Good positioning call — "fast judgment on conversion leaks" is a sharper read than "coaching." That actually maps to how the skill works: diagnostic, not advisory. Going to test that framing in the next copy pass. Appreciate it.
Exactly.
“Fast judgment on conversion leaks” is the product.
“AI Growth Coach” undersells it because coaching sounds ongoing, soft, and service-like.
What you’ve built sounds more like:
run the page
find the leak
fix what’s costing conversions
That is much sharper.
If this stays a Gumroad skill, the current name is fine.
But if you’re planning to turn it into a real diagnostic product, the name needs to carry more authority than “AI Growth Coach.”
That’s where Beryxa fits better.
It gives you room to build the product around conversion judgment instead of being boxed into the coach category.
Appreciate the push on this. The naming question is real — if this outgrows the Gumroad-skill format into a real diagnostic product, we'll re-test the name then. For now the framework's still calibrating against more pages. If anyone reading this wants their landing put through the 12-check pass, drop the URL here and we'll post the output back.
That makes sense.
If the framework is still calibrating, the right move is to prove the diagnostic is sharp first.
Once people start trusting the output, the name becomes much more important.
Right now the useful test is simple:
do users treat it like advice,
or do they treat it like evidence?
If they read the output and think “interesting feedback,” it stays a coach.
If they read it and think “this is costing conversions,” then it becomes a diagnostic product.
That’s the line I’d watch.
That's a sharp test. "Advice vs evidence" is the line I haven't been able to articulate cleanly until now.
What I'm watching for: when someone runs the 12-check on their landing page, do they push back on a finding ("hmm, not sure I agree with that one") or do they ship the rewrite ("that's actually broken, I need to fix it"). The first is advice. The second is evidence.
Right now the signal's mixed — about half each. Probably means the framework needs to be more confident on the high-impact issues (refuse to be hedgy where we shouldn't be) and more honest about uncertainty on the lower-impact ones, so the user doesn't have to do the discounting themselves.
If you've got a landing page you want me to run through it, happy to send the output back here as a public test of the framework. Either way — this is the kind of feedback that makes the product better, not the kind that just feels good.
That mixed signal tells you exactly where the product is right now.
The framework is strong enough to create trust sometimes, but not yet sharp enough to make every high-impact issue feel undeniable.
That’s the gap.
I’d separate the output into two layers:
high-confidence conversion leaks
and lower-confidence observations
The first should sound like evidence.
The second can sound like guidance.
If everything is framed the same way, users have to decide what matters.
But if the product clearly says “this is costing conversions,” it starts behaving less like a coach and more like a diagnostic system.
That’s also the point where the name starts mattering more.
Because if users begin treating it as evidence, “AI Growth Coach” will feel too soft for what the product is actually doing.
Two-layer output is the right call and we're shipping it in the next pass. Top tier — "this is costing conversions" — fires only on the 4-5 issues with hard pattern evidence (CTA verb mismatch, pricing-tier inversion, value-prop word count over threshold). Bottom tier — "consider this" — covers strong heuristics where context could legitimately override.
The discipline is in NOT promoting heuristics into the top tier. If we get that wrong, users learn to discount the whole output. Bar: would I bet $14 of credibility on this fix moving the needle on a page I've never seen?
On the name — agreed it's downstream. Rename trigger: when users start quoting top-tier findings as "the audit said this is costing conversions" rather than "the audit suggested..." That's the line.
Most useful product-design conversation we've had on the skill so far. Thanks.
One practical follow-up on this because the “advice vs evidence” distinction stuck with me.
If the next version is separating high-confidence conversion leaks from lower-confidence observations, then this is no longer just a copy issue. It becomes a trust and positioning issue.
The product needs users to read the output as diagnosis, not coaching.
That is where I think a focused naming/positioning audit would be useful before the next version gets more users: current name risk, whether “AI Growth Coach” is making people expect advice, how to frame the diagnostic layer, what language makes the audit feel credible, and when the name should change if users start treating the output as evidence.
Not a long consulting thing. Just a sharp written breakdown around the exact product transition you described.
I’m doing a few of these at $99 while refining the format.
If useful, connect here and I can give you a clear outside read before the next copy/product pass:
https://www.linkedin.com/in/aryan-y-0163b0278/
That rename trigger makes sense.
If users start quoting it as “the audit said this is costing conversions,” then the product has clearly moved from feedback into evidence.
The only thing I’d keep pressure-testing is whether the name affects that behavior earlier than expected.
“AI Growth Coach” naturally makes people expect advice, suggestions, and interpretation.
But the product you’re describing is trying to create judgment: hard-pattern evidence, confidence tiers, and conversion leaks worth acting on.
That means the name may not just be downstream. It may influence whether users read the output as coaching or diagnosis in the first place.
So I’d still watch the language carefully.
If people say “the coach suggested,” the current name is probably reinforcing the soft frame.
If they say “the audit found this leak,” then the product is earning a more serious category.
That is probably the moment to stop treating the name as cosmetic and start treating it as part of the trust system.