<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>tinkrbox blog</title>
    <link>https://tinkrbox.com/blog</link>
    <description>Practical articles on AI development, LLM security, legacy modernization, and AI governance in Japan, from the tinkrbox team.</description>
    <language>en</language>
    <item>
      <title>Demystifying CVE-2026-114938: The &quot;JavaToShell&quot; Exploit Chain and How to Secure Your Architecture</title>
      <link>https://tinkrbox.com/blog/javatoshell-cve</link>
      <guid>https://tinkrbox.com/blog/javatoshell-cve</guid>
      <pubDate>Wed, 18 Nov 2026 00:00:00 GMT</pubDate>
      <description>JavaToShell (CVE-2026-114938) is a CVSS 10.0 expression-language injection flaw in modern Java frameworks that hands unauthenticated attackers an OS shell before request validation ever runs. We break down the exploit chain, the risk to AI agent and RAG orchestration layers, and the patches, WAF rules, and architectural controls to apply now.</description>
    </item>
    <item>
      <title>Before you launch an LLM app: a security checklist</title>
      <link>https://tinkrbox.com/blog/llm-app-security-checklist</link>
      <guid>https://tinkrbox.com/blog/llm-app-security-checklist</guid>
      <pubDate>Wed, 18 Nov 2026 00:00:00 GMT</pubDate>
      <description>Prompt injection, data leaks, and over-powered agents are the risks that reach production. Here is what to test before your LLM feature goes live, and how.</description>
    </item>
  </channel>
</rss>
