Website Conversion Path Auditor

RevenuePath Guard renders a public commercial page in a real browser, identifies visible buyer-intent controls, and safely checks a bounded set of same-origin GET destinations. It returns structured evidence for placeholder CTAs, missing fragments, disabled controls, and failed routes.

The wedge is deliberately narrower than a broken-link crawler: it prioritizes signup, demo, pricing, buy, subscribe, checkout, quote, and account-creation paths instead of emitting every link on a site.

What the MVP does

• renders JavaScript-driven pages with Playwright;
• classifies visible commercial CTAs using deterministic rules;
• checks up to 6 unique same-origin routes with a pinned HEAD request and a byte-capped GET fallback only when HEAD is unsupported;
• records HTTP failures, disabled controls, missing fragments, and placeholder routes;
• captures tightly clipped CTA evidence screenshots only for candidate findings;
• emits structured JSON on Apify; the local CLI also writes a compact Markdown report;
• blocks credentials in URLs, private or reserved IP space, mixed public/private DNS answers, nonstandard ports, action-like paths, sensitive query parameters, and unsafe redirect targets;
• routes browser traffic through a DNS-pinning egress proxy, blocks non-read methods, popups, forms, downloads, WebSockets, beacons, service workers, media, and non-proxied WebRTC;
• checks robots.txt and stops conservatively when the target path is disallowed or the policy cannot be checked safely.

It does not click controls, submit forms, log in, create accounts, purchase, message, upload, download, or mutate the target. It is not a vulnerability scanner and it does not claim that a technical finding caused lost revenue. The MVP is limited to English-language, public, unauthenticated landing pages.

Local run

npm install
npm test
npm run audit -- --url https://example.com --output artifacts/example

Use --fail-on-high to return exit code 3 when a high-severity candidate failure is found. That makes the CLI usable in CI without turning medium-confidence findings into hard failures.

Apify Actor

The .actor directory contains bounded input and output schemas, a dataset view, and a Playwright Docker image. Under pay-per-event pricing, src/actor.mjs verifies that one completed-site-audit event fits the caller's limit, completes the audit, and couples the event charge to delivery of the structured dataset result.

The launch price is $3 per completed site audit, charged once only when the structured audit result is returned. Marketplace pricing is configured in Apify Console rather than hard-coded in this repository. External publication, payout verification, and later price changes remain owner-controlled actions.

Output health values

• action_needed: at least one high-severity candidate failure was reproduced.
• review: no high finding, but at least one medium finding needs review.
• clear_in_bounded_scan: no candidate failure was found within the declared limits; this is not a whole-site guarantee.