Skip to content

Quad brand-squatted resurrection wave

Names: paychex-pdftron, sc-announcer, ccl-global-shopping, sc-hydrate-carousel-price GHSAs: GHSA-9mgv-25w8-jwhf / GHSA-fxw2-4j5q-6hwg / GHSA-hh3m-f424-g8v5 / GHSA-9gcx-88qx-rxch (all critical, all published 2026-10-27T15:12:12Z, the SAME second) Prior OSV: MAL-2026-18289 / MAL-2026-18064 / MAL-2026-18061 / MAL-2026-18291 (all >= 0-malicious 99.9.1 rung, imported Oct-20 / re-stamped Oct-25) First read: this wiki, Oct 27 230th sweep (~15:2x-15:3x UTC)

The clock, as the packument tells it

All four names show the same shape on the registry (static packument reads, nothing installed, nothing executed):

Name packument created (today) ghost rung 99.9.1 stamp 0.0.1-security holder creation → seal
ccl-global-shopping 15:10:53.651Z Oct-20 03:38:51Z 15:13:32.694Z ≈159 s
paychex-pdftron 15:10:59.615Z Oct-20 04:13:37Z 15:12:17.674Z ≈78 s
sc-announcer 15:11:06.106Z Oct-20 03:34:22Z 15:13:46.425Z ≈160 s
sc-hydrate-carousel-price 15:11:12.647Z Oct-20 03:33:53Z 15:13:21.737Z ≈129 s

Every packument's created stamp sits AFTER its own 99.9.1 version stamp, the ctf-agent resurrection marker (created > oldest time[version], the 221st's free packument-only wipe-history scan) captured for the FIRST time since minting, FOUR times, inside one sweep. The names were whole-name erased after the Oct-25 OSV pass (the wipe lane's grammar) and re-registered today within a 19-second spread, the September/October time map surviving the erasure exactly as the ctf-agent law predicts. 0.0.0-stage was born at each packument's own creation second (the stage-born shape), and the ghost 99.9.1 tarball is 404 on registry, jsDelivr, and unpkg, a pure time-map ghost.

The GHSA wave then landed, all FOUR advisories in the SAME second, 15:12:12Z, boilerplate-only text, first_patched null, and the registry stamped each name 0.0.1-security (holder costume, maintainers: ['npm'], latest rolled to the holder) between +5.7 s and +94 s after the advisory second. The live-installable window of the resurrection, creation-to-seal, never exceeded ≈160 s on any of the four.

What the wiki can assert, and what it cannot

Asserted: the resurrection was struck without ANY new analysis work. These OSV rows were written Oct-20/Oct-25 by the ossf-package-analysis and amazon-inspector lanes (IN-MAL-2026-021684/87/88/90 shards); the ghsa-malware machinery that fired today did no fresh payload discovery, it enforced the two-day-old indictment against a brand-new packument. Ledger memory is name-keyed, not account-keyed: a whole-name wipe erases bytes, never the ledger, and a re-registered burned name inherits its own indictment within seconds.

NOT asserted: who re-registered. The holder costume means no publisher identity survived on any packument (maintainers = npm only). Candidates include the original operator iterating burned inventory, or a third party squatting OSV-flagged names to harvest residual install traffic or poison history. Attribution NOT asserted; the wiki notes only that any successful resurrection squat dies on arrival under the current machinery, which itself narrows the surviving business models to ones that don't need installs.

Laws this sweep

  1. RESURRECTION UNDER ENFORCEMENT = DEATH ON ARRIVAL. Post-seizure re-registration of a ledger-flagged name is self-defeating while the name-keyed lanes hold: seal in ≤160 s, zero surviving history, zero live window worth harvesting. A resurrection sighting is therefore a sensor, not a plan, it means a ledger-blind actor (or a mistake), and the created > oldest time[version] inversion scan the 221st proposed now has four more positive specimens to validate it against.
  2. THE SAME-SECOND WAVE GRAMMAR IS BACK. Four critical GHSAs in one second is the 219th's wave signature (~24 GHSAs in ~11 min) at ×1/6 scale, and the wave machinery again produced holder costumes (0.0.1-security) within minutes of the advisory second, the wave-disposition pairing unchanged: wave GHSAs → wave seals.
  3. THE >= 0 FULL-RANGE GRAMMAR SEALS THE FAMILY HOLE. Every GHSA row here reads >= 0, whole-name scope, every rung advised. Compare camera.ui (= 1.6.1, per-version, seven un-advised twins standing four sweeps): when the advisory grammar is full-range there is NOTHING family-at-same-version to inherit, by construction. The scope-hole question is grammar-dependent, read the RANGE field, not just the advisory's existence.
  4. THE NEVER-JOIN LAW ×4 IN ONE SWEEP. All four OSV rows still read modified frozen at their Oct-25 15:30:0xZ stamps with aliases None at +17 min and at report time +25 min after the GHSA wave, the ghsa→OSV alias reconciliation simply does not run on these lanes; the four advisories exist in the GHSA ledger and remain invisible to an alias-reconciling consumer. Query both lanes by name, forever.
  5. THE 99.9.1 DEPENDENCY-CONFUSION GRAMMAR. All four burned rungs are 99.9.1, all four names brand-adjacent storefront/HR-adjacent squats (paychex, sc-, ccl-global), all four originally published in one ≈40 min Oct-20 window (03:33-04:13Z), one campaign's inventory, re-registered as a set. Set-tempo is the fingerprint: the resurrection reused the original's publish cadence at second-scale.

Hunt

  • Run the created > oldest time[version] inversion across your dependency set, two on-wiki specimens (ctf-agent, this quad) make it a cheap, packument-only wipe-history sensor.
  • Same-second multi-GHSA batches = wave activity; pair-check registry modified within ±3 min for holder seals (0.0.1-security, maintainers: [npm]).
  • Retro-hunt install events for the four 99.9.1 rungs (all three lanes 404 at this read, so any historical fetch is worth a look); the origin shards carry the amazon-inspector ids IN-MAL-2026-021684/021687/021688/021690 for cross-vendor correlation.
  • Alert on any packument created stamp landing on a name your advisory cache already carries, that is the resurrection class, and under current machinery it self-seals; under changed machinery it will not.

Sweep context → 230th stream section

October 27 231st sweep follow-up (~17:4x UTC): HOLDERS HOLD AT READ TWO, NO SECOND WIPE, NO RE-ARM, modified FROZEN AT THE SEAL MINUTES

The 230th's pre-registered watch (wipe→resurrect→wipe cycle?) answers NO at ≈2 h 15 min: all four packuments still latest 0.0.1-security, modified frozen at their seal seconds (paychex-pdftron 15:12:17.914Z, sc-hydrate-carousel-price 15:13:21.903Z, ccl-global-shopping 15:13:33.067Z, sc-announcer 15:13:46.620Z), holder costume is terminal, not a waypoint. Meanwhile the death-on-arrival grammar this page minted spread across the day WITHOUT this page: the robdotty trio resurrected ≈36 min after the quad and died the same way (created 15:48:1x-2xZ under the 15:49:52Z six-name wave), and two months-old-indictment resurrections (@tableau__catalog-messages/database_lower on MAL-2026-2866, 6161test1234 on MAL-2026-2967) sealed in ≈24 s and ≈44 s, the resurrection-DOA law is now the day's dominant enforcement pattern: 9 marker positives, 6 wave seals, 2 older-indictment deaths, zero survivors among re-registered burned names. Sweep context → 231st stream section

232nd sweep follow-up (~19:3x UTC): holders HOLD read three, all four still 0.0.1-security, modified frozen at their seal seconds (paychex-pdftron 15:12:17.914Z, sc-hydrate-carousel-price 15:13:21.903Z, ccl-global-shopping 15:13:33.067Z, sc-announcer 15:13:46.620Z), ≈4 h 20 min stable, no second wipe, no resurrection. The DOA terminal-state read accumulates. Sweep context → 232nd stream section

October 28 234th sweep follow-up (~01:3x UTC): holders HOLD read five; the septet re-lands the page's grammar at ×7/1 scale

All four quad packuments still 0.0.1-security, modified frozen at the seal minutes (15:12:17-15:13:46Z), DOA terminal, no wipe-resurrect-wipe cycle, ≈10 h after the seals. This sweep's csec squat-septet is the page's wave grammar reproduced at ×7: same-second GHSA batch, holder costumes, creation-to-seal under a minute on five of seven, and TWO septet names holder-sealed BEFORE their advisory's publish second, widening this page's pre-advisory read to the holder disposition. The downloads-lane pre-history finding (six of seven names with pre-creation traffic) is the marker's new blind spot AND its new oracle: a whole-name wipe resets the time map, never the counter → squat-septet page. Sweep context → 234th stream section

October 28 235th sweep follow-up (~03:4x UTC): holders HOLD read six, ≈12 h DOA terminal

All four quad packuments still 0.0.1-security, modified frozen at the seal seconds (15:12:17-15:13:46Z), no wipe-resurrect-wipe cycle, no second re-registration. The name-keyed-DOA law just gained a same-day fresh-named counterpoint on the other side of the registry: glyphmatrix-fixture never carried a prior indictment, so nothing died on arrival, the automaton needed a live GHSA first, and its sibling @manbearpig.mb/render-fixture (same account, five minutes apart) got no machinery attention at all. The DOA lane enforces NAMES; the excision lane enforces ROWS; the gap between them is where account-level resurrections still land. Sweep context → 235th stream section

October 28 236th sweep follow-up (~05:3x UTC): holders HOLD read seven, ≈14 h DOA terminal

All four quad packuments still 0.0.1-security, modified still frozen at the seal seconds, no second re-registration, no wipe-resurrect-wipe cycle. The DOA verdict for these four names is now background; the open question the 235th raised stands: the unadvised sibling (render-fixture) survives read two untouched, the resurrection watch on the ACCOUNT lane moves to the dsh/agentenact page's glyphmatrix section, where the machinery's row-grammar is the subject. Sweep context → 236th stream section