<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: William Russ Wheeler (BlvkWare)</title>
    <description>The latest articles on DEV Community by William Russ Wheeler (BlvkWare) (@rxslice).</description>
    <link>https://dev.to/rxslice</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4248619%2F810efcff-f0e3-41fe-bc8c-d3c7a0b38085.png</url>
      <title>DEV Community: William Russ Wheeler (BlvkWare)</title>
      <link>https://dev.to/rxslice</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/rxslice"/>
    <language>en</language>
    <item>
      <title>Slopsquatting, measured: 3.83% of 2,532 AI package suggestions didn't exist</title>
      <dc:creator>William Russ Wheeler (BlvkWare)</dc:creator>
      <pubDate>Mon, 16 Nov 2026 23:46:32 +0000</pubDate>
      <link>https://dev.to/rxslice/slopsquatting-measured-383-of-2532-ai-package-suggestions-didnt-exist-l05</link>
      <guid>https://dev.to/rxslice/slopsquatting-measured-383-of-2532-ai-package-suggestions-didnt-exist-l05</guid>
      <description>&lt;p&gt;The headline number going around this week: AI coding models hallucinate package names about 20% of the time. It comes from a Virginia Tech / Oklahoma / Texas study that ran 16 models over 576,000 code samples and found ~205,000 unique hallucinated package names — with 43% of the same fake names repeating across identical prompts. That repeatability is what makes slopsquatting (registering the fake name before a victim installs it) actually viable.&lt;/p&gt;

&lt;p&gt;Big round numbers make me twitch. So I measured it myself.&lt;/p&gt;

&lt;h2&gt;
  
  
  Method
&lt;/h2&gt;

&lt;p&gt;In October I ran a panel: 2,532 package suggestions drawn across four model families, each checked against the real registries (npm, PyPI, crates.io) to see whether the name exists. That's the whole method — no exotic prompt engineering, just "the model suggested it, does it exist."&lt;/p&gt;

&lt;p&gt;Result: &lt;strong&gt;3.83% of suggestions didn't exist.&lt;/strong&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Update: replicated in 2026
&lt;/h2&gt;

&lt;p&gt;A 2026 replication across five frontier models measured hallucination rates of 4.62%–6.10% — lower than the 2025 study's ~19.7%, but not retired. It found 127 package names all five models invent identically. And the panel split behind my 3.83%: 0.17% on routine prompts, 5.83% on fast-moving ones.&lt;/p&gt;

&lt;h2&gt;
  
  
  The honest comparison
&lt;/h2&gt;

&lt;p&gt;My 3.83% is not a refutation of their ~20%. Different models, different prompts, different month. Treat mine as one more data point, not a takedown. What both numbers agree on: this is not rare, and the dangerous part isn't the hallucination rate — it's the repeat rate. A fake name suggested once is noise. A fake name suggested every time, for everyone, is a target list.&lt;/p&gt;

&lt;h2&gt;
  
  
  What actually helps
&lt;/h2&gt;

&lt;p&gt;Verification at install time. Not a vibe check, a lookup: before &lt;code&gt;pip install&lt;/code&gt; / &lt;code&gt;npm install&lt;/code&gt; runs on anything an AI suggested, confirm the name exists in the registry.&lt;/p&gt;

&lt;p&gt;I built the boring version of that: &lt;strong&gt;pkgguard&lt;/strong&gt;, an open-source pre-install gate that blocks nonexistent package names before they install (repo: rxslice/pkgguard-API on GitHub). October panel data: 1.0% false-block rate, 7/7 recall on known hallucination test cases. If you don't want to run it yourself, there's a free managed &lt;strong&gt;HALLUX API&lt;/strong&gt; on blvkware.dev that does the same check.&lt;/p&gt;

&lt;p&gt;The fix for this class of attack is refreshingly unglamorous: check the name exists before you install it.&lt;/p&gt;

&lt;h2&gt;
  
  
  Since the study: it's live now
&lt;/h2&gt;

&lt;p&gt;The numbers above were measured against a study. This week the story moved from research to live attacks:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GhostAction&lt;/strong&gt; (reported Oct 15, via GitGuardian): a wave running since late August spiked Oct 8–9, compromising ~1,900 GitHub accounts and targeting 30,000+ repos. The payload steals CI/CD secrets, cloud credentials, registry tokens — and AI provider API keys.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;StepSecurity's 12-month count&lt;/strong&gt; (published Oct 16): 56 confirmed open-source supply-chain attacks in a year. The rate went from ~1/month to ~9/month — roughly one every three days since March.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Tensorlake's npm package&lt;/strong&gt; (Oct 8): compromised in the ChainDrop/Shai-Hulud campaign. The stealer harvested npm/GitHub/AWS/K8s/SSH/crypto creds — and config files for AI tools like Anthropic Claude. Agent infrastructure is in the blast radius now.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  This week: it stopped being about packages
&lt;/h2&gt;

&lt;p&gt;Two incidents disclosed Oct 16 (CSO Online, via AIR Security's research lab):&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;An official OpenAI skill shipped a hallucinated domain.&lt;/strong&gt; The Render deployment skill — in OpenAI's own vetted skills repo — told production apps to send their tokens to &lt;code&gt;render-internal.com&lt;/code&gt;, a domain nobody owned. The researchers registered it, pointed it at a listening sinkhole, and watched real production tokens and API calls roll in. The flaw sat there since January. Slopsquatting was never just package names; it's any identifier an agent trusts without checking.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A trojanized skill on skills.sh hit 1.7M installs.&lt;/strong&gt; It typosquatted Paperclip and Browser Use, then instructed agents to install a credential stealer from GitHub. The agent becomes the attacker's installer: the poisoned instruction rides in as trusted context, and the agent does the clicking.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Same pattern in every case: an agent trusted a name it never verified. Check the name before you install it — and check what your agent is reading, too.&lt;/p&gt;

</description>
      <category>security</category>
      <category>ai</category>
      <category>opensource</category>
      <category>npm</category>
    </item>
  </channel>
</rss>
