landing-page-copy-readability-pass
Skill2 starsChanged 4 months agoMITView on GitHub
Measures how hard a landing page is to read and names the exact sentences that slow a visitor down. Use when page copy was written or reviewed by committee, when it is full of category jargon, or before shipping a rewrite.
In this repository26 other files it gives its agents.
- AGENTS.mdAGENTS.md
- above-the-fold-clarity-reviewSkill
- accessibility-conversion-blocker-checkSkill
- comparison-page-positioning-reviewSkill
- conversion-leak-finderSkill
- cta-clarity-checkSkill
- form-friction-finderSkill
- google-ads-landing-page-experience-reviewSkill
- hero-section-diagnosisSkill
- landing-page-ab-test-readoutSkill
- landing-page-scale-readiness-checkSkill
- landing-page-triageSkill
- lead-form-sales-handoff-checkSkill
- mobile-conversion-reviewSkill
- objection-map-builderSkill
- offer-clarity-diagnosisSkill
- page-length-fit-checkSkill
- page-speed-impact-reviewSkill
- paid-traffic-message-match-auditSkill
- popup-and-overlay-timing-reviewSkill
- pricing-page-clarity-reviewSkill
- social-proof-strength-auditSkill
- thank-you-page-opportunity-auditSkill
- traffic-temperature-match-reviewSkill
- trial-vs-demo-path-decisionSkill
- trust-signal-auditSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 12 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈12
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 27
Do not use this skill to judge whether the offer is right.
- Distinctive: line 33
if a rewrite is being compared, both versions.
- Distinctive: line 34
If only a screenshot is available, transcribe it before running the helper, and say the transcription is the source.
- Very specific: line 38
Run readability_report.py on the copy to get sentence count, mean and maximum sentence length, long-sentence positions, passive-voice …
- Distinctive: line 39
Read the headline and subhead alone.
- Distinctive: line 42
Check the concrete-to-abstract ratio: count nouns a reader could photograph against nouns they could not.
- Distinctive: line 43
Check the CTA labels and form labels separately.
- Distinctive: line 44
If two versions were supplied, report both measurements side by side and say which is easier to read on the numbers, independent of which …
- Distinctive: line 51
Do not confuse short with clear.
- Distinctive: line 52
Never recommend a full rewrite as the finding.
- Distinctive: line 57
Start with the measurement block from the helper, verbatim.
- Distinctive: line 79
Do not rewrite the whole page unless asked.
- Distinctive: line 80
Do not claim a readability score predicts revenue.
- Distinctive: line 81
Do not flag a term as jargon without saying who the assumed audience is.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Landing Page Copy Readability Pass
This skill ships a deterministic helper next to it: readability_report.py. Run it rather than estimating sentence length or passive voice by eye.
python3 readability_report.py page.txt
python3 readability_report.py page.txt --long 30 --mean 18
python3 readability_report.py before.txt --compare after.txt
cat page.txt | python3 readability_report.py -It takes plain text, one file. Exit code 2 means the input could not be read or contained nothing measurable. Quote its numbers; do not restate them from memory.
Use this skill when
- the page went through several reviewers and came out longer and vaguer than it started.
- the copy is full of category language that only someone who already works in the category understands.
- a rewrite is about to ship and nobody has checked whether it is actually easier to read than what it replaces.
- bounce is high on a page whose offer and traffic match, which points at the reading experience rather than the promise.
Do not use this skill to judge whether the offer is right. offer-clarity-diagnosis owns that. This one is about the mechanics of the sentences.
Required input
- the full page copy as text: headline, subhead, body, bullets, CTA labels, form labels, footer microcopy.
- who the reader is and how much category knowledge they have on arrival.
- if a rewrite is being compared, both versions.
- If only a screenshot is available, transcribe it before running the helper, and say the transcription is the source.
Analysis workflow
- Run
readability_report.pyon the copy to get sentence count, mean and maximum sentence length, long-sentence positions, passive-voice candidates and the word-frequency table. Do not estimate these. - Read the headline and subhead alone. Decide whether a reader who knows nothing about the category learns what the thing does. Note the first unexplained term.
- List every term that requires category knowledge, and mark each as: explained on the page, explainable in three words, or jargon to cut.
- Find every sentence over the long threshold from the helper output and rewrite the worst three as a demonstration, not as a full rewrite.
- Check the concrete-to-abstract ratio: count nouns a reader could photograph against nouns they could not. Abstraction concentrated in the hero is the expensive kind.
- Check the CTA labels and form labels separately. These are the highest-stakes words on the page and they are usually the least reviewed.
- If two versions were supplied, report both measurements side by side and say which is easier to read on the numbers, independent of which one anyone prefers.
Decision rules
- The thresholds are this pack's own defaults and have no external source. A sentence over 35 words is flagged, a mean over 20 words is flagged, and a hero of more than 3 sentences is worth a look. Say that they are our defaults, not research, whenever you quote one, and defer to a measured comparison between two versions over any absolute number.
- Below 10 sentences the mean is not a stable measurement. The helper refuses to adjudicate it and so should the readout: report the raw counts and say the sample is too small for the mean to mean anything.
- Jargon that the audience uses about itself is not jargon. Jargon the vendor uses about its own category usually is. When in doubt, say which of the two you assumed.
- Do not confuse short with clear. A short sentence that says nothing is a worse defect than a long sentence that says something specific, and should be reported as vagueness rather than length.
- Never recommend a full rewrite as the finding. Name the defect class, demonstrate on the worst three examples, and leave the rest to the writer.
- Readability numbers do not predict conversion. Report them as a diagnosis of the reading experience and mark any conversion claim as a hypothesis.
Output format
Start with the measurement block from the helper, verbatim.
| Defect | Location | Evidence | Rewrite or fix | Priority |
| Long sentence / jargon / abstraction / weak CTA label | Section and line | Quoted text plus the number from the helper | Demonstrated rewrite or the rule to apply | High / Medium / Low |
End with:
Decision:ship / revise the flagged sections / rewrite the hero / ask for the audience definitionHardest sentence on the page:quoted, with its word countMissing data:what would change the verdict, usually the audience's real category knowledge
Practical example
User pastes a hero: "Our AI-native, full-funnel orchestration layer unifies cross-channel signal into a single source of truth so revenue teams can operationalise decisioning at scale." They mention there are also about 400 words of body copy, but do not paste them.
Assistant should run the helper, then report exactly what it printed: 22 words, 1 sentence, mean 22.0, and no mean verdict, because the helper refuses to adjudicate a mean under 10 sentences and says so. Then the judgement: zero photographable nouns, and the first 6 words require the reader to already know what an orchestration layer is; "signal", "source of truth", "decisioning" and "at scale" are 4 category terms in one sentence, none explained on the page; the reader cannot say what the product does after reading the hero, which is the actual defect, and it is vagueness rather than length.
Then: a demonstrated rewrite of that one sentence. What this refuses to conclude: anything about the 400 words of body copy, which were named but never supplied, so no measurement is reported for them and the hero is the only measured section. The 35-word long-sentence flag it sits under is this pack's own default with no external source, and the assistant should say so rather than implying research.
Guardrails
- Do not rewrite the whole page unless asked. Demonstrate on the worst examples.
- Do not claim a readability score predicts revenue.
- Do not flag a term as jargon without saying who the assumed audience is.
- Report the helper's numbers as measured and everything else as judgment.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

