ghost-cursor

---

What ghost-cursor actually does

Generates Bézier-curve mouse paths with human-plausible acceleration for Puppeteer, addressing the behavioural layer rather than the static fingerprint one.

What ghost-cursor's maintenance signals mean

The liveness field here is pushed_at, read as 2026-08-11 against a field check 3 days later. That is inside 90 days, and 90 days is where the second edition of the rubric put ACTIVE after moving it down from the first edition's 180.

The reason for the move is worth stating, because it is the kind of threshold that looks arbitrary until you know what it is measuring against: detection changes on a shorter cycle than half a year, so a six-month-old evasion is not a current one. What a recent push does not tell you is what was in it. A commit landing on a branch is not a release, not a fix, and not a test — it is evidence that somebody is still there.

No release was captured for this entry, which this register records as unknown rather than as none. The distinction matters more than it looks: a project with no tags and a project whose tags the sweep failed to read produce the same blank cell, and printing them as the same fact would be the register asserting something it has not measured.

Where ghost-cursor sits in this catalogue

ghost-cursor is filed in section 02, automation stealth plugins, patches & cdp-leak fixes, whose scope is this: Patches and plugins for Puppeteer, Playwright and Selenium, plus the proxies and cursor libraries that sit beside them. An entry sits in exactly one section here. Where a project could defensibly go in two, it is placed where a reader would look for it first rather than where it fits most tidily, and that is an editorial decision rather than a measured one.

Filed beside it in the same section: CloudflareBypassForScraping (ACTIVE), hazetunnel (NRND), HumanCursor (NRND). Comparing an entry against its own section-mates is more informative than comparing it against the catalogue as a whole, because everything in a section is answering roughly the same question and the differences left over are the ones that matter.

It is published under MIT, alongside the rest of the MIT entries in this register, and it sits with everything else currently marked ACTIVE. Read those two facets together: a permissively licensed project that stopped shipping is not a safer dependency than a copyleft one that did not, and a filter on one axis alone will tell you the wrong thing.

Substitutes for the same job, with no claim that either is better: HumanCursor. The alternative relation is symmetric and carries no ranking, because this register publishes no ranking of any kind. The full ghost-cursor comparison table sets all of them side by side on licence, push date, release cadence and state.

What to settle before you adopt ghost-cursor

Behavioural signals are only worth investing in once the static ones are consistent; a perfect mouse path on a contradictory fingerprint does not help. Check whether your target actually collects pointer telemetry before adding the dependency.