- Name
-
@di-atomic/cro-audit
- Tagline
-
Why your page isn't converting, with a benchmark behind every answer.
- Hero subhead
-
cro-audit reads a live page and diagnoses it across 8 conversion dimensions, and every finding cites a benchmark, so you get evidence instead of one more opinion about your hero. As of v0.2.0 it also re-runs a funnel audit, so you can find out whether the last set of findings ever got fixed.
- Who it's for
-
For anyone with traffic that doesn't convert, and a pile of contradictory advice about why.
- Version
-
0.2.0
- Tier
-
beta
- Path type
-
guidance-only
- Hero image
-
- Install command
-
opvs-skills install @di-atomic/cro-audit
- Trigger phrase
-
audit this landing page for conversion
- Marketplace URL
-
https://market.opvs.ai/v/di-atomic/cro-audit
- Blog URL
-
https://di-atomic.com/blog/shipping-di-atomic-cro-audit
- Stat 1 value
-
100/100
- Stat 1 label
-
marketplace security score
- Stat 2 value
-
8
- Stat 2 label
-
conversion dimensions scored on every audit
- Stat 3 value
-
5
- Stat 3 label
-
scripts that gate the audit's own output
- Stat 4 value
-
60
- Stat 4 label
-
benchmark cards behind the findings
- What headline
-
What you actually get
- Cap 1 icon
-
📊
- Cap 1 title
-
8 dimensions, not a hunch
- Cap 1 desc
-
Clarity, proof, friction, hierarchy, urgency, trust, mobile and measurement all get scored, so a weak page cannot hide behind the one dimension nobody checked.
- Cap 2 icon
-
🔍
- Cap 2 title
-
Every finding cites a benchmark
- Cap 2 desc
-
Each finding names a specific element on your page and the measured figure behind the call. A platitude linter fails the audit if it produces advice you could have written without reading the page.
- Cap 3 icon
-
🔁
- Cap 3 title
-
Re-run it without burning a test account
- Cap 3 desc
-
Split the funnel where it writes, not where the pages change. Everything upstream of that boundary re-runs at zero cost, so checking last quarter's findings costs you nothing.
- Worked title
-
One untouched URL, four readings
- Worked example
-
The same page, unchanged, returned an LCP of 3032, 1100, 848 and 2272 ms across four sequential runs. A single sample would have reported it fixed. A verdict now needs the samples to agree, or it returns UNKNOWN.
- Worked framing
-
Why a re-run reports three states, not two
- Rel 1 icon
-
🎯
- Rel 1 name
-
@di-atomic/cro-actions
- Rel 1 desc
-
Turns the ranked finding list into a sequenced plan where every action names a component to install or a specific edit.
- Rel 1 url
-
https://di-atomic.com/skills/cro-actions
- Rel 2 icon
-
📈
- Rel 2 name
-
@di-atomic/cro-optimizer
- Rel 2 desc
-
Closes the loop after a fix ships: decides whether the result is real and reverts confirmed losers.
- Rel 2 url
-
https://di-atomic.com/skills/cro-optimizer
- Rel 3 icon
-
✍
- Rel 3 name
-
@di-atomic/copy-engine
- Rel 3 desc
-
Writes the replacement copy once the audit has named which element is costing you the conversion.
- Rel 3 url
-
https://di-atomic.com/skills/copy-engine
- CTA headline
-
Find out why the page leaks, with evidence.
- CTA description
-
Install the skill, point it at the page, and get a ranked finding list where every line names an element and cites a number.
- Security score
-
100
