Multiply the daily npm rate by 365 and the twelve months to May 2026 produced on the order of 6,570 catalogued malicious packages in a single ecosystem [4][18]. That figure is why the GitHub engineer who leads the Dependabot team, writing at Help Net Security, calls per-record human review theater. It cannot scale. It also relocates alert fatigue from users into an internal queue that quietly backs up [19][15].
The control sits somewhere else. Upstream sources publish mistakes, including reports that land on the wrong package name and bulk imports that go sideways overnight [11]. His requirement is that every ingested record keeps a trail back to the exact upstream change that produced it, that imports revert as a batch in one move in minutes, and that provenance is treated as an operational control on the same shelf as feature flags and rollbacks [10]. He also cites a cap on how many advisories a single input can generate, though the text available breaks off mid-sentence before naming the unit [16].
Eight ecosystems is where the judgement calls pile up. Registries disagree on whether package names are case-sensitive. Affected versions arrive as exact pins in one format and open-ended ranges in another, and the boundary semantics differ just enough to hurt. Severity vocabularies do not line up either [9]. A consumer inherits all of that: the database was built by someone else's processes, shaped by someone else's judgement calls, and the automation simply acts on what it is handed [20].
GitHub sits on both sides of that exchange. It had been flagging npm malware since March, per the author, before extending coverage [3], which makes it a contributor to the commons it now reads from, and community repositories strip the context of who found what first [12].
The author argues the same lessons hold whether the input is package malware reports, OSV records, or ISAC indicators [17]. That generalization is the part worth carrying: joining a feed is procurement. Acting on one is incident response conducted on someone else's evidence.
