Website Conversion Audit Skill
Turns a subjective "this page feels off" into a scored, prioritised, measurable action list. Built from the buyer's point of view: the only question that matters is what stands between a qualified visitor and the action you want them to take.
When to use
- Before a redesign, so effort goes to the findings that actually move revenue.
- When demo requests or sign-ups are lower than traffic justifies.
- To turn an informal landing-page review into a ranked, defensible plan.
When NOT to use
- For pure performance/Core Web Vitals work — use Lighthouse/PageSpeed; this skill assumes the page loads fine and looks at the funnel.
- For content/SEO ranking — this is about converting the visitor who already arrived.
Method: five conversion pillars
Every finding maps to one pillar and is scored by revenue impact, not effort:
- Clarity — can a buyer tell what this is and who it is for within 5 seconds?
- CTA — is the primary action obvious, singular, and low-friction?
- Form — does the form ask only for what is needed, and can marketing route/score the lead?
- Trust — is proof (logos, numbers, security) visible before the ask?
- Flow — broken links, hidden assets, mobile breakage, dead ends.
Severity
- P0 — directly blocks or leaks conversions right now (broken CTA, hidden primary asset, form that can't be routed). Fix this week.
- P1 — meaningful drag on conversion (weak hero, buried proof). Fix this month.
- P2 — polish that compounds over time.
Inputs
Required:
url — the page auditedobservations — structured findings from a manual pass (see scripts/audit_score.py for the schema)
Output (JSON)
A verdict line, a pillar_scores map, and findings[] sorted most-severe first, each with pillar, severity, what, why, fix, measure.
Run it
python scripts/audit_score.py # runs the built-in sample page
python scripts/audit_score.py page.json # audits your own observations file
Runs with zero dependencies and no API keys. The scoring logic is deterministic and inspectable in scripts/audit_score.py.
1---2name: website-conversion-audit3description: Audits a B2B website page for conversion leaks from a buyer's lens and returns findings ranked by revenue impact. Takes a URL plus a set of observed page elements (hero message, CTAs, form fields, trust signals, mobile behaviour) and returns P0/P1/P2 findings, each with what was found, why it costs conversions, a concrete fix, and how to measure the fix. Use before a redesign, when diagnosing why demo requests are low, or to turn a landing page review into a prioritised action list.4---56# Website Conversion Audit Skill78Turns a subjective "this page feels off" into a scored, prioritised, measurable action list. Built from the buyer's point of view: the only question that matters is what stands between a qualified visitor and the action you want them to take.910## When to use1112- Before a redesign, so effort goes to the findings that actually move revenue.13- When demo requests or sign-ups are lower than traffic justifies.14- To turn an informal landing-page review into a ranked, defensible plan.1516## When NOT to use1718- For pure performance/Core Web Vitals work — use Lighthouse/PageSpeed; this skill assumes the page loads fine and looks at the funnel.19- For content/SEO ranking — this is about converting the visitor who already arrived.2021## Method: five conversion pillars2223Every finding maps to one pillar and is scored by revenue impact, not effort:24251. **Clarity** — can a buyer tell what this is and who it is for within 5 seconds?262. **CTA** — is the primary action obvious, singular, and low-friction?273. **Form** — does the form ask only for what is needed, and can marketing route/score the lead?284. **Trust** — is proof (logos, numbers, security) visible before the ask?295. **Flow** — broken links, hidden assets, mobile breakage, dead ends.3031## Severity3233- **P0** — directly blocks or leaks conversions right now (broken CTA, hidden primary asset, form that can't be routed). Fix this week.34- **P1** — meaningful drag on conversion (weak hero, buried proof). Fix this month.35- **P2** — polish that compounds over time.3637## Inputs3839Required:40- `url` — the page audited41- `observations` — structured findings from a manual pass (see `scripts/audit_score.py` for the schema)4243## Output (JSON)4445A `verdict` line, a `pillar_scores` map, and `findings[]` sorted most-severe first, each with `pillar`, `severity`, `what`, `why`, `fix`, `measure`.4647## Run it4849```bash50python scripts/audit_score.py # runs the built-in sample page51python scripts/audit_score.py page.json # audits your own observations file52```5354Runs with zero dependencies and no API keys. The scoring logic is deterministic and inspectable in `scripts/audit_score.py`.
saurabhshuklagrowisto