<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>RADAR — patch_now</title>
    <link>https://radar.trulysovereign.com</link>
    <atom:link href="https://radar.trulysovereign.com/feeds/action/patch-now.xml" rel="self" type="application/rss+xml"/>
    <description>Items whose recommended action is patch_now.</description>
    <language>en</language>
    <lastBuildDate>Sat, 21 Nov 2026 19:19:18 GMT</lastBuildDate>
    <item>
      <title>AIT-GUI route chain allows unauthenticated Python execution</title>
      <link>https://radar.trulysovereign.com/items/342293b5-dd6a-480f-957e-bae6fb2c5452</link>
      <guid isPermaLink="false">342293b5-dd6a-480f-957e-bae6fb2c5452</guid>
      <pubDate>Sat, 21 Nov 2026 19:19:11 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-jrjx-9jjw-747m on 20 November 2026, rating CVE-2026-77309 high severity. The advisory states that a route chain in ait-gui allows unauthenticated Python execution in the server process. Version 2.5.1 patches the vulnerability.

Why it matters: An unauthenticated attacker can execute arbitrary Python code in the server process, achieving full remote code execution without credentials. Any exposed ait-gui instance running an unpatched version is compromised on contact.

What to do: Upgrade ait-gui to version 2.5.1 or later. If you cannot upgrade immediately, take the service offline or restrict network access to authenticated internal users only until patched.</description>
    </item>
    <item>
      <title>GKE patches CVE-2026-99288 for Container-Optimized OS 1.32 and 1.33</title>
      <link>https://radar.trulysovereign.com/items/0ad8c973-9170-4755-9319-f592f756d3f2</link>
      <guid isPermaLink="false">0ad8c973-9170-4755-9319-f592f756d3f2</guid>
      <pubDate>Sat, 21 Nov 2026 19:17:51 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://docs.cloud.google.com/support/bulletins/index#gcp-2026-088">Google Cloud security bulletins</source>
      <description>Google added GKE patch versions for Container-Optimized OS minor versions 1.32 and 1.33 to bulletin GCP-2026-088 on 2026-11-20. The patches address CVE-2026-99288, a Linux kernel TCP denial-of-service vulnerability that can cause kernel panic. Patches for GKE 1.37 on Container-Optimized OS and Ubuntu node images remain in progress.

Why it matters: The vulnerability can crash affected virtual machines through a kernel panic. GKE clusters running Container-Optimized OS on versions 1.32 or 1.33 are now patchable; clusters on 1.37 or Ubuntu remain exposed until those patches release.

What to do: Remediation: see bulletin https://docs.cloud.google.com/support/bulletins/index#gcp-2026-088.</description>
    </item>
    <item>
      <title>pip prior to v26.2.2 bundles compromised eMudhra certificates</title>
      <link>https://radar.trulysovereign.com/items/9dd24bbc-1f5b-4a07-885f-9467aaa0fd0d</link>
      <guid isPermaLink="false">9dd24bbc-1f5b-4a07-885f-9467aaa0fd0d</guid>
      <pubDate>Sat, 21 Nov 2026 07:10:51 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://mail.python.org/archives/list/security-announce@python.org/message/4234WU4ONSVWPT3BIGG6TOJO2UFE26Y6/">Python security announcements</source>
      <description>pip versions before v26.2.2 include certifi with compromised eMudhra certificates. The vulnerability is rated CRITICAL severity and tracked as CVE-2026-133949.

Why it matters: Compromised certificate authorities can issue fraudulent certificates for any domain, enabling man-in-the-middle attacks on HTTPS connections made by pip during package installation. Any system using affected pip versions is vulnerable when downloading packages.

What to do: Upgrade pip to v26.2.2 or later. If upgrading pip directly, first upgrade Python to a version that rejects the compromised intermediates (CVE-2026-115557), or use hash pinning with a known-secure pip version during the upgrade process.</description>
    </item>
    <item>
      <title>GitLab 19.6.1, 19.5.8, 19.4.10 patch CVE-2026-132825 heap overflow in Ruby</title>
      <link>https://radar.trulysovereign.com/items/4dc3d6cc-44c4-442c-9006-1100c3224920</link>
      <guid isPermaLink="false">4dc3d6cc-44c4-442c-9006-1100c3224920</guid>
      <pubDate>Sat, 21 Nov 2026 07:10:43 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-6-1-released/">GitLab patch releases</source>
      <description>GitLab released versions 19.6.1, 19.5.8, and 19.4.10 on November 20, 2026, patching CVE-2026-132825, a heap-based buffer overflow in Ruby's string case conversion. The vulnerability affects all GitLab CE/EE versions before 19.4.10, 19.5 before 19.5.8, and 19.6 before 19.6.1. The Ruby security team states the flaw may allow remote code execution in some scenarios. GitLab.com and GitLab Dedicated are already patched.

Why it matters: A heap overflow with potential RCE exposure is present in the Ruby runtime underlying every unpatched self-managed GitLab installation. Self-compiled installations require rebuilding Ruby with version 3.3.14 or 3.4.13, not just updating GitLab packages.

What to do: Upgrade self-managed GitLab installations to 19.6.1, 19.5.8, or 19.4.10 immediately. For self-compiled installations, rebuild the Ruby installation with Ruby 3.3.14 or 3.4.13 before or during the GitLab upgrade. Official GitLab Linux packages and container images already include the patched Ruby.</description>
    </item>
    <item>
      <title>NGINX HTTP/2 vulnerabilities enable DoS and potential RCE</title>
      <link>https://radar.trulysovereign.com/items/24c15b5a-7949-468d-bd4d-f3576f81e225</link>
      <guid isPermaLink="false">24c15b5a-7949-468d-bd4d-f3576f81e225</guid>
      <pubDate>Fri, 20 Nov 2026 20:23:17 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://docs.cloud.google.com/support/bulletins/index#gcp-2026-089">Google Cloud security bulletins</source>
      <description>F5 disclosed CVE-2026-123633 and CVE-2026-123007 on 18 November 2026, affecting NGINX Open Source 1.19.7-1.31.8 and 1.9.5-1.31.8 respectively, and NGINX Plus R33+. Under specific conditions, an unauthenticated remote attacker can crash NGINX worker processes or potentially achieve remote code execution. CVE-2026-123633 requires reverse proxying and worker connection depletion; CVE-2026-123007 affects HTTP/2 traffic including static content serving. Google Cloud managed services (Cloud Load Balancing, GKE control plane, Cloud Run, Cloud Functions, App Engine) are not affected.

Why it matters: Self-managed NGINX deployments on GCP — GKE Ingress controllers, Compute Engine VMs, and Marketplace images — are exposed to remote DoS and potential RCE if serving HTTP/2 traffic. The RCE risk increases if ASLR is disabled or bypassed. Google has delisted affected Click-to-Deploy NGINX Marketplace images until patched versions are available.

What to do: Upgrade NGINX Open Source to 1.31.9 (mainline) or 1.30.7 (stable), or NGINX Plus to 37.0.7.1/37.2.0.1 (for CVE-2026-123633) or 37.0.8.1/37.3.0.1 (for CVE-2026-123007) per F5 advisory K000164058. If immediate patching is not possible, disable HTTP/2 on exposed NGINX listeners (set use-http2: &quot;false&quot; in ingress-nginx or remove http2 listener directives) or place a Google Cloud External Application Load Balancer configured for HTTP/1.1 to backends in front of NGINX. Note that disabling HTTP/2 does not mitigate the eight additional November 18 F5 CVEs, so upgrading remains necessary.</description>
    </item>
    <item>
      <title>Mistune 3.3.3 and earlier: quadratic-time DoS in inline image parsing</title>
      <link>https://radar.trulysovereign.com/items/d35d0e29-a6c2-4032-989a-540b5d979eab</link>
      <guid isPermaLink="false">d35d0e29-a6c2-4032-989a-540b5d979eab</guid>
      <pubDate>Fri, 20 Nov 2026 07:21:49 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-ww7q-mxrj-5ffg on 19 November 2026, rating a denial-of-service vulnerability in mistune (CVE-2026-97699) as high severity. The flaw is in inline image parsing (precedence_scan) and bypasses not-is-image guards. First patched version: 3.3.4.

Why it matters: Mistune is a markdown parser in Python dependency trees. An attacker can trigger quadratic-time processing, causing the parser to hang or exhaust resources on crafted input.

What to do: Check whether mistune is in your dependency tree and whether you are running 3.3.3 or earlier. If so, update to 3.3.4.</description>
    </item>
    <item>
      <title>MCP Python SDK could send OAuth credentials to attacker-chosen server</title>
      <link>https://radar.trulysovereign.com/items/1add7a7a-43d4-432a-97b6-fe55e3b5c3d5</link>
      <guid isPermaLink="false">1add7a7a-43d4-432a-97b6-fe55e3b5c3d5</guid>
      <pubDate>Fri, 20 Nov 2026 07:21:35 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-qx49-fqc8-xw99 (CVE-2026-115563) on 19 November 2026, rating it high severity. The Model Context Protocol Python SDK mcp package allows an MCP server to choose the OAuth authorization server, enabling credential interception. Patched in mcp 1.30.0 and 2.2.0.

Why it matters: An application using the vulnerable SDK to authenticate with an MCP server can be tricked into sending OAuth credentials to an attacker-controlled authorization endpoint. The attacker gains the credentials and can impersonate the application.

What to do: Check whether you depend on mcp (Model Context Protocol Python SDK) at any version below 1.30.0 or 2.x below 2.2.0. If so, update to 1.30.0, 2.2.0 or later in this cycle.</description>
    </item>
    <item>
      <title>Mistune 3.3.3 and earlier vulnerable to quadratic-time DoS</title>
      <link>https://radar.trulysovereign.com/items/e5e47794-943e-4277-afed-734499be073f</link>
      <guid isPermaLink="false">e5e47794-943e-4277-afed-734499be073f</guid>
      <pubDate>Fri, 20 Nov 2026 07:21:26 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-phmp-hrw4-r36r on 19 November 2026, rating CVE-2026-97698 high severity. Mistune versions before 3.3.4 are vulnerable to quadratic-time denial of service through unbounded parenthesis scanning when parsing empty-label links. Version 3.3.4 patches the vulnerability.

Why it matters: An attacker can cause CPU exhaustion in any service that parses untrusted Markdown through mistune, including user-generated content, documentation sites, and API endpoints that accept Markdown input. The quadratic complexity means processing time grows exponentially with input size.

What to do: Upgrade mistune to 3.3.4. Check your dependency tree with `pip list | grep mistune` or `poetry show mistune` to confirm the installed version, as mistune is a transitive dependency in many Python Markdown toolchains.</description>
    </item>
    <item>
      <title>NLTK 3.10.3 and earlier allows local code execution via prover9/mace4 config</title>
      <link>https://radar.trulysovereign.com/items/fd177730-0f17-41d1-b548-32de03af008f</link>
      <guid isPermaLink="false">fd177730-0f17-41d1-b548-32de03af008f</guid>
      <pubDate>Fri, 20 Nov 2026 07:21:12 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-cc5r-64rf-75hg on 19 November 2026, rating CVE-2026-129234 high severity. NLTK versions before 3.10.4 use an untrusted search path in prover9/mace4 configuration, allowing local code execution through a relative binary_location value. Version 3.10.4 patches the vulnerability.

Why it matters: An attacker with local access can execute arbitrary code by placing a malicious binary in the search path before NLTK resolves the prover9 or mace4 binary location. This affects any deployment where NLTK's theorem proving components run with untrusted local users or in shared environments.

What to do: Upgrade nltk to 3.10.4 or later. If you cannot upgrade immediately, verify whether your code calls NLTK's prover9 or mace4 interfaces and restrict local access to systems running those components.</description>
    </item>
    <item>
      <title>django-docutils before 0.30.0 allows file disclosure, SSRF, and XSS</title>
      <link>https://radar.trulysovereign.com/items/9e8c70e0-7d41-4c19-914c-cee0aabc6d7f</link>
      <guid isPermaLink="false">9e8c70e0-7d41-4c19-914c-cee0aabc6d7f</guid>
      <pubDate>Fri, 20 Nov 2026 07:20:59 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-86mf-q4h5-69gp (CVE-2026-59988) on 20 November 2026, rating django-docutils high severity. Versions before 0.30.0 use unsafe Docutils defaults that enable local file disclosure, server-side request forgery, and stored cross-site scripting. Version 0.30.0 patches the vulnerability.

Why it matters: Applications using django-docutils to render user-supplied reStructuredText can leak local files to attackers, make requests to internal services, or store malicious scripts that execute in other users' browsers. All three attack vectors are exploitable through the same unsafe configuration.

What to do: Upgrade django-docutils to 0.30.0 or later. Check whether your application accepts reStructuredText input from users and passes it to django-docutils rendering functions.</description>
    </item>
    <item>
      <title>flyto-core 2.26.8 patches SSRF in http.paginate Link header handling</title>
      <link>https://radar.trulysovereign.com/items/c1de4d06-f963-45d7-a558-8543f14a55e6</link>
      <guid isPermaLink="false">c1de4d06-f963-45d7-a558-8543f14a55e6</guid>
      <pubDate>Fri, 20 Nov 2026 07:20:46 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>CVE-2026-77315 published 20 November 2026. flyto-core's http.paginate follows server-supplied next-page URLs from Link headers or cursors without revalidating the destination, enabling SSRF. First patched version is 2.26.8.

Why it matters: An attacker controlling a paginated API response can redirect http.paginate to internal endpoints, cloud metadata services, or other restricted targets your application can reach but should not query on external instruction.

What to do: Upgrade flyto-core to 2.26.8 or later. If you use http.paginate with untrusted APIs, verify you are running a vulnerable version before 2.26.8 and prioritize the update.</description>
    </item>
    <item>
      <title>Jupyter MCP Server unauthenticated routes allow backend hijack in standalone HTTP mode</title>
      <link>https://radar.trulysovereign.com/items/16bd8fa7-0e06-4bda-ab4a-7f60d0b992fb</link>
      <guid isPermaLink="false">16bd8fa7-0e06-4bda-ab4a-7f60d0b992fb</guid>
      <pubDate>Fri, 20 Nov 2026 07:20:37 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-24pw-rf5p-mgp7 (CVE-2026-77318) on 20 November 2026, rating it high severity. Jupyter MCP Server before version 1.0.3 exposes unauthenticated management routes that can hijack the Jupyter backend when deployed in standalone HTTP mode.

Why it matters: An attacker with network access to the management routes can take control of the Jupyter backend without authentication, compromising notebook execution environments and any data or credentials accessible to the runtime.

What to do: Update jupyter-mcp-server to version 1.0.3. If you run Jupyter MCP Server in standalone HTTP mode, verify whether management routes are exposed to untrusted networks and prioritize the update.</description>
    </item>
    <item>
      <title>CodeWhale auto_approve defaults to true, bypassing task-path hardening</title>
      <link>https://radar.trulysovereign.com/items/152a381f-f439-4071-acb0-3f82400d19b0</link>
      <guid isPermaLink="false">152a381f-f439-4071-acb0-3f82400d19b0</guid>
      <pubDate>Fri, 20 Nov 2026 07:20:26 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>GitHub published GHSA-8hp3-xxrg-v9c8 (CVE-2026-83817) on 19 November 2026, rating it high severity. The advisory states that CodeWhale's automation auto_approve parameter defaults to true, which bypasses the task-path hardening fix intended to prevent silent recurring agent execution. Patched in version 0.9.0 for both codewhale and codewhale-tui packages.

Why it matters: The default configuration allows agents to execute tasks without approval, circumventing security controls designed to prevent unauthorized recurring execution. The triage notes indicate an active abuse pattern involving silent recurring agent execution.

What to do: Upgrade codewhale and codewhale-tui to version 0.9.0. If you cannot upgrade immediately, explicitly set auto_approve to false in automation configurations to restore the intended approval gate.</description>
    </item>
    <item>
      <title>Effect 3.22.1 patches ReDoS in Schema.TemplateLiteral regex</title>
      <link>https://radar.trulysovereign.com/items/0b1127fa-f36c-4e64-b6bc-be7431dc8609</link>
      <guid isPermaLink="false">0b1127fa-f36c-4e64-b6bc-be7431dc8609</guid>
      <pubDate>Fri, 20 Nov 2026 07:20:17 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>CVE-2026-121707 discloses a regular expression denial of service vulnerability in the effect npm package's Schema.TemplateLiteral compiled regex. GitHub rated it high severity. Effect 3.22.1 patches the issue, published 19 November 2026.

Why it matters: An attacker who controls input to Schema.TemplateLiteral can cause the regex engine to hang, blocking the event loop and denying service to other requests in the same process.

What to do: Update effect to 3.22.1 or later if you use Schema.TemplateLiteral with untrusted input. Check your dependency tree for transitive uses of effect.</description>
    </item>
    <item>
      <title>CodeWhale TUI exec_shell classifier enables silent RCE via grep flags</title>
      <link>https://radar.trulysovereign.com/items/c5c5abc6-5a16-4cd6-8fbd-beedf7fa6c86</link>
      <guid isPermaLink="false">c5c5abc6-5a16-4cd6-8fbd-beedf7fa6c86</guid>
      <pubDate>Fri, 20 Nov 2026 07:20:08 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>GitHub published GHSA-3mgh-93r5-r7jf (CVE-2026-83816) on 19 November 2026. CodeWhale TUI versions before 0.9.0 auto-approve exec flags in fd/rg/git grep commands through the parallel-readonly classifier, enabling remote code execution. Severity: high. Patched in version 0.9.0.

Why it matters: An attacker can execute arbitrary code by injecting malicious flags into grep operations that the classifier incorrectly treats as read-only. The 'silent RCE' designation indicates exploitation leaves minimal traces.

What to do: Upgrade codewhale-tui and codewhale packages to version 0.9.0. Grep your dependency trees for 'codewhale-tui' and 'codewhale' to identify affected installations.</description>
    </item>
    <item>
      <title>CVE-2026-99288 kernel panic in GKE via crafted TCP packets</title>
      <link>https://radar.trulysovereign.com/items/a8b2921f-a32e-4faf-b9cb-9c1d5d7517ef</link>
      <guid isPermaLink="false">a8b2921f-a32e-4faf-b9cb-9c1d5d7517ef</guid>
      <pubDate>Fri, 20 Nov 2026 07:18:29 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://docs.cloud.google.com/support/bulletins/index#gcp-2026-088">Google Cloud security bulletins</source>
      <description>Google published GCP-2026-088 on 2026-11-19 for CVE-2026-99288, a denial-of-service vulnerability in the Linux kernel TCP implementation affecting GKE Standard and Autopilot clusters running Container-Optimized OS or Ubuntu. A remote attacker can trigger a kernel panic by sending specially crafted TCP packets to any reachable TCP listener on a node, Pod, or control plane IP endpoint. Internal VPC traffic can also trigger it.

Why it matters: Both control planes and worker nodes are vulnerable to remote denial-of-service from any network position that can reach a TCP listener, including from within the same VPC. The cluster control plane itself is a reachable target. Affected nodes will kernel panic, taking down running workloads.

What to do: Check whether your GKE clusters run Container-Optimized OS or Ubuntu (gVisor and Windows Server nodes are not affected). Remediation: see bulletin https://docs.cloud.google.com/support/bulletins/index#gcp-2026-088. The full article could not be retrieved (attempted 2026-11-20T07:18:20.598Z); this item uses the feed entry only.</description>
    </item>
    <item>
      <title>OpenID4VP accepts forged X.509 certificates reusing trusted anchor Subject DNs</title>
      <link>https://radar.trulysovereign.com/items/abfdb2ae-5c3e-459f-86b8-81f2b746e237</link>
      <guid isPermaLink="false">abfdb2ae-5c3e-459f-86b8-81f2b746e237</guid>
      <pubDate>Thu, 19 Nov 2026 20:40:35 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>CVE-2026-75522 published 19 November 2026 for @openeudi/openid4vp. The library's X.509 chain validation accepts a forged certificate that reuses a trusted anchor's Subject Distinguished Name without verifying the signature. Patched in version 0.8.1.

Why it matters: An attacker can present a certificate that appears to chain to a trusted root by copying its Subject DN, bypassing signature verification. This breaks the trust model for any OpenID4VP credential verification that relies on X.509 certificate chains.

What to do: Update @openeudi/openid4vp to 0.8.1 or later. Verify exposure first if you are unsure whether your application uses this library for credential verification.</description>
    </item>
    <item>
      <title>install-artifact-from-github loads native artifacts without integrity verification</title>
      <link>https://radar.trulysovereign.com/items/4676d530-87a0-4e89-b079-4f4ab5e81184</link>
      <guid isPermaLink="false">4676d530-87a0-4e89-b079-4f4ab5e81184</guid>
      <pubDate>Thu, 19 Nov 2026 07:30:38 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>CVE-2026-73864 published 18 November 2026 for install-artifact-from-github. The package loads downloaded native artifacts without verifying integrity. Severity rated high. First patched version is 1.7.0.

Why it matters: An attacker who can intercept or modify the download can substitute malicious native code that executes with the privileges of the build or installation process. No integrity check prevents this.

What to do: Update install-artifact-from-github to 1.7.0 or later. Check package.json and lock files for any version below 1.7.0.</description>
    </item>
    <item>
      <title>Command injection in @augment-vir/node grep() via unsanitized searchPattern</title>
      <link>https://radar.trulysovereign.com/items/3c752202-1632-4b5c-9a75-81fc70ebaa8c</link>
      <guid isPermaLink="false">3c752202-1632-4b5c-9a75-81fc70ebaa8c</guid>
      <pubDate>Wed, 18 Nov 2026 20:36:28 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>GitHub published GHSA-xvcc-vqjv-phqh on 18 November 2026, a high-severity command injection vulnerability (CVE-2026-63496) in @augment-vir/node. The grep() function does not sanitize the searchPattern parameter. Version 31.73.5 patches the issue.

Why it matters: An attacker who controls the searchPattern argument can execute arbitrary shell commands on the host. Any code path that passes user input or untrusted data to grep() is exploitable.

What to do: Update @augment-vir/node to 31.73.5 or later. Grep your codebase for imports of @augment-vir/node and calls to grep() to confirm exposure before patching.</description>
    </item>
    <item>
      <title>GCP-2026-078 updated with GKE patch versions for CVE-2026-98432</title>
      <link>https://radar.trulysovereign.com/items/5a72809d-808b-43f0-a41f-0b10be98ed3b</link>
      <guid isPermaLink="false">5a72809d-808b-43f0-a41f-0b10be98ed3b</guid>
      <pubDate>Wed, 18 Nov 2026 20:34:58 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://docs.cloud.google.com/support/bulletins/index#gcp-2026-078">Google Cloud security bulletins</source>
      <description>Google Cloud security bulletin GCP-2026-078 was updated on 2026-11-18 to add patch versions for GKE. The bulletin covers CVE-2026-98432, a Linux kernel IPv4 TCP stack vulnerability that can cause kernel panic or container escape with root code execution on the node, exploitable by an unprivileged process.

Why it matters: GKE clusters running unpatched versions are exposed to container escape attacks. An attacker with unprivileged access inside a container can execute code as root on the underlying node, breaking container isolation and potentially compromising the entire cluster.

What to do: Check your GKE cluster versions against the patched versions listed in the bulletin. Remediation: see bulletin https://docs.cloud.google.com/support/bulletins/index#gcp-2026-078 Remediation: see bulletin https://docs.cloud.google.com/support/bulletins/index#gcp-2026-078.</description>
    </item>
    <item>
      <title>libp2p gossipsub creates unbounded topics map on distinct subscriptions</title>
      <link>https://radar.trulysovereign.com/items/1edbed37-889c-4563-9745-537704453026</link>
      <guid isPermaLink="false">1edbed37-889c-4563-9745-537704453026</guid>
      <pubDate>Wed, 18 Nov 2026 07:32:26 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>CVE-2026-119793 published 17 November 2026 for @libp2p/gossipsub. The default configuration creates a permanent topics-map entry for each distinct subscription with no cap, rated high severity. First patched version is 17.0.0.

Why it matters: An attacker can exhaust memory by sending subscriptions with unique topic names, causing denial of service. The map grows without bound in the default configuration.

What to do: Upgrade @libp2p/gossipsub to 17.0.0 or later. If you cannot upgrade immediately, verify whether your application accepts gossipsub subscriptions from untrusted peers.</description>
    </item>
    <item>
      <title>@libp2p/crypto ASN.1 decoder stalls Node event loop on attacker input</title>
      <link>https://radar.trulysovereign.com/items/5f836177-4bb2-4594-830c-3469f1a2924c</link>
      <guid isPermaLink="false">5f836177-4bb2-4594-830c-3469f1a2924c</guid>
      <pubDate>Wed, 18 Nov 2026 07:32:16 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>GitHub published GHSA-rvfr-cqwq-rc4j on 17 November 2026. CVE-2026-110789, rated high severity. The ASN.1 DER decoder in @libp2p/crypto drives an unbounded synchronous loop on attacker-controlled length fields, stalling the Node.js event loop before authentication. Patched in version 5.1.22.

Why it matters: An unauthenticated attacker can freeze the event loop by sending crafted input, denying service to all connections handled by that process. The vulnerability is pre-auth, so no credentials or prior access are required.

What to do: Update @libp2p/crypto to 5.1.22 or later. Check your dependency tree for transitive uses of this package — networking libraries commonly pull it in.</description>
    </item>
    <item>
      <title>mail-parser 4.6.0 and earlier vulnerable to ReDoS in header parsing</title>
      <link>https://radar.trulysovereign.com/items/bba1e449-40f1-4d48-ac81-24e3dcaa2c15</link>
      <guid isPermaLink="false">bba1e449-40f1-4d48-ac81-24e3dcaa2c15</guid>
      <pubDate>Tue, 17 Nov 2026 07:57:10 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-r7w7-vpq8-c5vv on 16 November 2026, rating mail-parser vulnerable to polynomial regular expression denial of service (ReDoS) during email header parsing. CVE-2026-129295 assigned. First patched version is 4.6.1.

Why it matters: An attacker can cause CPU exhaustion by sending crafted email headers, blocking the parser and any service that depends on it. Services that parse untrusted email are exposed.

What to do: Check whether you run mail-parser 4.6.0 or earlier. If so, update to 4.6.1.</description>
    </item>
    <item>
      <title>mail-parser 4.6.2 patches sender IP spoofing in get_server_ipaddress()</title>
      <link>https://radar.trulysovereign.com/items/954ceb2c-8653-418a-bf57-a371d981eec4</link>
      <guid isPermaLink="false">954ceb2c-8653-418a-bf57-a371d981eec4</guid>
      <pubDate>Tue, 17 Nov 2026 07:57:02 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-72px-j8gh-p79x on 16 November 2026, a high-severity advisory for CVE-2026-129297 in mail-parser. The vulnerability allows sender IP address spoofing in the get_server_ipaddress() function. Version 4.6.2 patches the issue.

Why it matters: Applications using mail-parser to extract or validate sender IP addresses can be deceived about message origin, bypassing IP-based access controls, allowlists, or audit trails that depend on accurate sender identification.

What to do: Check whether mail-parser is in your dependency tree and upgrade to 4.6.2 or later if you are running an affected version.</description>
    </item>
    <item>
      <title>code-review-mcp-server command injection patched in 1.5.3</title>
      <link>https://radar.trulysovereign.com/items/4e3060ca-5962-4da9-a601-16f9f5dee17b</link>
      <guid isPermaLink="false">4e3060ca-5962-4da9-a601-16f9f5dee17b</guid>
      <pubDate>Tue, 17 Nov 2026 07:56:51 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=npm&amp;severity=high&amp;per_page=100">GitHub advisories — npm</source>
      <description>GitHub published GHSA-jq3q-62jc-fqq9 on 16 November 2026, documenting a high-severity command injection vulnerability (CVE-2026-68934) in code-review-mcp-server caused by unsanitized folderPath and baseBranch parameters. Version 1.5.3 patches the issue.

Why it matters: An attacker who controls folderPath or baseBranch input can execute arbitrary commands on the host running the MCP server, compromising the system and any repositories or credentials it accesses.

What to do: Upgrade code-review-mcp-server to 1.5.3 or later. Check package.json and lock files for this dependency in developer tooling and CI pipelines.</description>
    </item>
    <item>
      <title>GitLab 19.5.6, 19.4.8, 19.3.9 patch 40 security vulnerabilities including high-severity CI/CD authorization bypass</title>
      <link>https://radar.trulysovereign.com/items/fe82f0dc-26f5-4f9d-b1bf-2df40bd65a66</link>
      <guid isPermaLink="false">fe82f0dc-26f5-4f9d-b1bf-2df40bd65a66</guid>
      <pubDate>Tue, 17 Nov 2026 07:56:21 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-5-6-released/">GitLab patch releases</source>
      <description>GitLab released versions 19.5.6, 19.4.8, and 19.3.9 on 16 November 2026, patching 40 security vulnerabilities. Two high-severity issues (CVSS 8.7) are XSS in wiki pages (CVE-2026-105739, affects 17.2+) and error tracking (CVE-2026-108153, affects 16.1+). CVE-2026-18241 (CVSS 8.2) is a race condition in CI/CD pipeline job authorization allowing access to protected variables and resources (affects 9.3+). CVE-2026-97187 (CVSS 7.6) allows reading Kubernetes secrets from other namespaces via workspaces (affects 17.6+). Affected versions: CE/EE 19.5 before 19.5.6, 19.4 before 19.4.8, 19.3 before 19.3.9, with some vulnerabilities reaching back to version 8.0.

Why it matters: The CI/CD authorization race condition exposes protected secrets and resources to authenticated users who should not have access. The workspace vulnerability leaks Kubernetes secrets across namespace boundaries in agent-connected clusters. XSS vulnerabilities allow arbitrary JavaScript execution in authenticated users' browsers. Multiple authorization bypasses expose private repository content, job artifacts, package metadata, and CI/CD templates to unauthorized users. Self-managed installations running any affected version are exposed.

What to do: Upgrade self-managed GitLab CE/EE installations to 19.5.6, 19.4.8, or 19.3.9 immediately. GitLab.com is already patched; GitLab Dedicated requires no action. Verify your current version against the impacted ranges listed for each CVE. No migrations are included; multi-node deployments should not require downtime.</description>
    </item>
    <item>
      <title>aiohttp 3.14.4 patches DoS via URL-encoded form fields</title>
      <link>https://radar.trulysovereign.com/items/bfa1a038-ce29-48c4-86d9-55b5f5136f6d</link>
      <guid isPermaLink="false">bfa1a038-ce29-48c4-86d9-55b5f5136f6d</guid>
      <pubDate>Mon, 16 Nov 2026 22:03:32 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-xgw8-8ccx-rmqh on 16 November 2026, rating CVE-2026-128126 high severity. The vulnerability affects aiohttp versions before 3.14.4: too many URL-encoded form fields block the event loop, causing denial of service. Version 3.14.4 patches the issue.

Why it matters: An attacker can block the event loop by sending requests with excessive URL-encoded form fields, making the service unresponsive. Any Python service using aiohttp to handle untrusted HTTP requests is exposed.

What to do: Upgrade aiohttp to 3.14.4 or later. Check requirements files and deployed containers for aiohttp &lt; 3.14.4.</description>
    </item>
    <item>
      <title>aiohttp unbounded buffering in upgrade requests</title>
      <link>https://radar.trulysovereign.com/items/71d95938-2aad-4118-9ba3-094311c7e1e1</link>
      <guid isPermaLink="false">71d95938-2aad-4118-9ba3-094311c7e1e1</guid>
      <pubDate>Mon, 16 Nov 2026 22:03:23 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-pw86-h8fm-j323 (CVE-2026-128129) on 16 November 2026, rating a high-severity unbounded buffering vulnerability in aiohttp during upgrade request handling. First patched version is 3.14.4.

Why it matters: Unbounded buffering during request handling allows an attacker to exhaust memory by sending crafted upgrade requests, causing service degradation or denial of service in applications using affected aiohttp versions.

What to do: Check your dependency tree for aiohttp versions below 3.14.4. If present, update to 3.14.4 or later in this cycle.</description>
    </item>
    <item>
      <title>aiohttp 3.14.4 patches event loop stall in wildcard matching</title>
      <link>https://radar.trulysovereign.com/items/36a458c5-a4a6-40e8-9292-ad50776a866f</link>
      <guid isPermaLink="false">36a458c5-a4a6-40e8-9292-ad50776a866f</guid>
      <pubDate>Mon, 16 Nov 2026 22:03:15 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-jm3w-gw2p-xjfj on 16 November 2026, a high-severity advisory for CVE-2026-128124 in aiohttp. The vulnerability causes event loop stalls in add_domain() wildcard matching. First patched version is 3.14.4.

Why it matters: An attacker can stall the async event loop, blocking all concurrent requests and responses in the application. This is a denial of service vector in the core async runtime that affects any service using aiohttp's domain matching.

What to do: Upgrade aiohttp to 3.14.4 or later. Check requirements files and lock files for aiohttp dependencies below 3.14.4.</description>
    </item>
    <item>
      <title>aiohttp 3.14.4 patches DoS via concatenated compressed streams</title>
      <link>https://radar.trulysovereign.com/items/e73106cd-74dd-41fe-8f8f-216b7eb36cd5</link>
      <guid isPermaLink="false">e73106cd-74dd-41fe-8f8f-216b7eb36cd5</guid>
      <pubDate>Mon, 16 Nov 2026 22:03:05 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published GHSA-h4cc-hm5h-fmm9 (CVE-2026-128131) on 16 November 2026, rating it high severity. aiohttp versions before 3.14.4 consume excessive CPU when processing concatenated compressed streams, creating a denial-of-service vector.

Why it matters: Services using aiohttp to handle HTTP requests can be forced into CPU exhaustion by a crafted payload, degrading or stopping availability. This affects any deployment where aiohttp processes untrusted input.

What to do: Upgrade aiohttp to 3.14.4 or later. Grep dependency manifests for aiohttp and check running versions against the patched release.</description>
    </item>
    <item>
      <title>NLTK dot2img enables local code execution via PATH manipulation</title>
      <link>https://radar.trulysovereign.com/items/a631b2fa-80eb-4fa3-b831-7833c05b3ee6</link>
      <guid isPermaLink="false">a631b2fa-80eb-4fa3-b831-7833c05b3ee6</guid>
      <pubDate>Mon, 16 Nov 2026 22:02:57 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>CVE-2026-128134 published 16 November 2026. NLTK's dot2img function runs Graphviz 'dot' without the trusted-exec chokepoint, allowing local code execution when an attacker controls a writable PATH element. First patched version: 3.10.4.

Why it matters: An attacker with write access to any directory in PATH can place a malicious 'dot' executable that NLTK will run with the application's privileges when dot2img is called. This bypasses the intended security boundary for external tool invocation.

What to do: Upgrade nltk to 3.10.4 or later. If you cannot upgrade immediately, audit PATH for writable directories and remove write permissions for untrusted users.</description>
    </item>
    <item>
      <title>aiohttp WebSocket receive queue limits bypassed by small messages</title>
      <link>https://radar.trulysovereign.com/items/6ee38546-5eb5-4f44-9d98-70cc88e7c05f</link>
      <guid isPermaLink="false">6ee38546-5eb5-4f44-9d98-70cc88e7c05f</guid>
      <pubDate>Mon, 16 Nov 2026 22:02:49 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>CVE-2026-128132 published 16 November 2026 as GHSA-5rmc-48h2-hvcm, rated high severity. Small or compressed WebSocket messages bypass receive queue limits in aiohttp. First patched version is 3.14.4.

Why it matters: An attacker can exhaust server resources by sending WebSocket messages that evade queue limits, causing denial of service. Applications using aiohttp WebSocket handling are exposed to resource exhaustion attacks.

What to do: Upgrade aiohttp to 3.14.4 or later. Grep for 'aiohttp' in requirements files and check deployed versions against the CVE scope.</description>
    </item>
    <item>
      <title>aiohttp WebSocket integer overflow bypasses max_msg_size</title>
      <link>https://radar.trulysovereign.com/items/63dec2da-4d8f-45bc-a8cc-24ba6bac1984</link>
      <guid isPermaLink="false">63dec2da-4d8f-45bc-a8cc-24ba6bac1984</guid>
      <pubDate>Mon, 16 Nov 2026 22:02:41 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>CVE-2026-128128 published 16 November 2026 for aiohttp, severity high. An integer overflow in WebSocket size validation allows messages to bypass max_msg_size limits. Patched in version 3.14.4.

Why it matters: Services using aiohttp WebSockets can receive messages larger than configured limits, potentially causing memory exhaustion or denial of service. The bypass is in protocol-level validation, so application-layer size checks do not prevent it.

What to do: Update aiohttp to 3.14.4 or later. If you run WebSocket endpoints with aiohttp, prioritize this in the current patch cycle.</description>
    </item>
    <item>
      <title>aiohttp 3.14.4 patches request replay and buffer growth vulnerability</title>
      <link>https://radar.trulysovereign.com/items/039ddc48-d9bc-430f-b22d-68eb94f840a9</link>
      <guid isPermaLink="false">039ddc48-d9bc-430f-b22d-68eb94f840a9</guid>
      <pubDate>Mon, 16 Nov 2026 22:02:32 GMT</pubDate>
      <category>patch_now</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>GitHub published CVE-2026-128130 on 16 November 2026, a high-severity vulnerability in aiohttp's pure-Python parser that allows request replay and buffer growth behind a declined upgrade. First patched version is 3.14.4.

Why it matters: Request replay can bypass authentication or rate limiting by re-executing requests the server already declined. Unbounded buffer growth consumes memory until the process fails or the host runs out of resources.

What to do: Check your dependency tree for aiohttp versions below 3.14.4. If present, update to 3.14.4 or later in this cycle.</description>
    </item>
    <item>
      <title>[CORRECTED] dnspython CVE-2026-128089 causes excessive memory and CPU use</title>
      <link>https://radar.trulysovereign.com/items/ce2fd1aa-e455-44fb-8cec-215077e7c30a</link>
      <guid isPermaLink="false">ce2fd1aa-e455-44fb-8cec-215077e7c30a</guid>
      <pubDate>Sun, 15 Nov 2026 06:36:05 GMT</pubDate>
      <category>patch_now</category>
      <category>corrected</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>[CORRECTED] GitHub published GHSA-vvf4-9x63-8xvv on 13 November 2026, rating CVE-2026-128089 in dnspython as high severity. The vulnerability is in zone-file parsing: the `from_text()` parser does not limit a WKS record service number to 0–65535, so a maliciously crafted zone file can cause excessive memory and CPU consumption. Versions earlier than 2.9.2 are affected. First patched version is 2.9.2.

Why it matters: Applications are affected only when they parse zone-file text whose record type and rdata are attacker-controlled. Using dnspython only for DNS resolution, without parsing attacker-controlled zone-file text, does not match the advisory's stated impact. A high severity rating does not mean every deployment that includes the package is exposed.

What to do: If your code parses zone-file format with dnspython (including `dns.rdata.from_text`) where an attacker can control the record type and rdata, upgrade to version 2.9.2 or later. If you only use dnspython as a DNS resolver and do not parse untrusted zone-file text, this advisory does not describe a demonstrated exposure, though upgrading remains ordinary hygiene.

Original text (as first published, superseded):

WHAT CHANGED: GitHub published GHSA-vvf4-9x63-8xvv on 13 November 2026, rating CVE-2026-128089 in dnspython as high severity. The vulnerability causes excessive memory and CPU consumption. First patched version is 2.9.2.

WHY IT MATTERS: Services using dnspython for DNS resolution can experience availability degradation or denial of service through resource exhaustion. The high severity rating indicates exploitability affects production workloads.

WHAT TO DO: Upgrade dnspython to version 2.9.2 or later in services where it is a runtime dependency.</description>
    </item>
    <item>
      <title>[CORRECTED] Neptune deserialization flaw lets authenticated users trigger unsafe query processing</title>
      <link>https://radar.trulysovereign.com/items/826a54ad-f6b8-4d18-8dde-5e0b8053ae57</link>
      <guid isPermaLink="false">826a54ad-f6b8-4d18-8dde-5e0b8053ae57</guid>
      <pubDate>Thu, 15 Oct 2026 17:05:00 GMT</pubDate>
      <category>patch_now</category>
      <category>corrected</category>
      <source url="https://aws.amazon.com/security/security-bulletins/rss/2026-136-aws/">AWS security bulletins</source>
      <description>[CORRECTED] AWS disclosed CVE-2026-110812 on 15 October 2026, a deserialization vulnerability in Apache TinkerPop's Gremlin functionality used by Amazon Neptune. All Neptune engine versions &gt;=1.2.0.0 are affected. An authenticated principal with Neptune data-plane query permissions can submit requests that cause the query engine to process untrusted input unsafely, potentially leading to unintended activity within the query engine environment.

Why it matters: AWS has fixed this in Neptune engine versions 1.4.7.1, 1.4.8.1 and 1.5.0.0. Clusters with auto minor version upgrade enabled get the fix in their next maintenance window; if you manage engine versions manually, you must upgrade the cluster yourself. An attacker who already has Neptune query permissions could cause the query engine to process untrusted input unsafely. The broad version range (1.2.0.0 and later) means most production Neptune clusters are affected.

What to do: Check your Neptune cluster engine versions in the AWS console. If running 1.2.0.0 or later, upgrade to the patched version for your engine line (1.4.7.1, 1.4.8.1 or 1.5.0.0), or confirm auto minor version upgrade is enabled and the fix lands in your next maintenance window. Until then, review IAM policies so only trusted principals have neptune-db data-plane (query) permissions, since exploitation requires authenticated query access.

Original text (as first published, superseded):

WHY IT MATTERS: This is a managed service vulnerability you cannot patch yourself — AWS must upgrade the engine. An attacker who already has Neptune query permissions can exploit this to execute unintended operations in the query engine environment, which in a managed service context means activity outside your direct control. The broad version range means most production Neptune clusters are affected.

WHAT TO DO: Check your Neptune cluster engine versions in the AWS console. If running &gt;=1.2.0.0, schedule an upgrade to the patched engine version listed in the full bulletin. Review IAM policies to confirm only necessary principals have neptune-db:* data-plane permissions, since exploitation requires authenticated query access.</description>
    </item>
    <item>
      <title>[CORRECTED] GitLab 19.3.5 replaces broken 19.3.4 patch release</title>
      <link>https://radar.trulysovereign.com/items/1df8a326-2157-425a-bad9-595ae9641187</link>
      <guid isPermaLink="false">1df8a326-2157-425a-bad9-595ae9641187</guid>
      <pubDate>Tue, 13 Oct 2026 02:46:44 GMT</pubDate>
      <category>patch_now</category>
      <category>corrected</category>
      <source url="https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-4-2-released/">GitLab patch releases</source>
      <description>[CORRECTED] GitLab released 19.3.5 on October 12, 2026, replacing 19.3.4 which had a bug affecting merge request pages when the Ona integration is enabled. All CVE impact statements now list 19.3.5 as the fixed version instead of 19.3.4. The release addresses 24 CVEs including CVE-2026-97229 (CVSS 10.0, unauthenticated RCE via CI/CD regex), CVE-2026-92883 (CVSS 8.7, XSS in syntax highlighting), CVE-2026-16301 (CVSS 8.5, unauthorized protected branch pipeline execution), and CVE-2026-14814 (CVSS 8.2, cache poisoning exposing CI/CD secrets).

Why it matters: If you patched to 19.3.4, merge request pages are broken for any installation using the Ona integration. If you have not patched yet, you are running a version with a critical unauthenticated remote code execution vulnerability (CVSS 10.0) that allows arbitrary code execution via a crafted CI/CD regex, plus multiple high-severity issues exposing secrets and allowing unauthorized pipeline execution.

What to do: If you are on 19.3.4 with Ona integration enabled, upgrade to 19.3.5 immediately to restore merge request functionality. If you are on any 19.3.x version before 19.3.5, upgrade to 19.3.5 to patch the critical RCE and other vulnerabilities. Self-managed installations on 19.2.x or 19.4.x should upgrade to 19.2.8 or 19.4.2 respectively.

Original text (as first published, superseded):

WHAT CHANGED: GitLab released 19.3.5 on October 12, 2026, replacing 19.3.4 which had a bug affecting merge request pages when the Ona integration is enabled. All CVE impact statements now list 19.3.5 as the fixed version instead of 19.3.4. The release addresses 26 CVEs including CVE-2026-97229 (CVSS 10.0, unauthenticated RCE via CI/CD regex), CVE-2026-92883 (CVSS 8.7, XSS in syntax highlighting), CVE-2026-16301 (CVSS 8.5, unauthorized protected branch pipeline execution), and CVE-2026-14814 (CVSS 8.2, cache poisoning exposing CI/CD secrets).</description>
    </item>
    <item>
      <title>[CORRECTED] PyJWT 2.13.0 accepts JWK public keys as HMAC secrets</title>
      <link>https://radar.trulysovereign.com/items/50c172c6-74a0-4090-a626-6f6df1236e26</link>
      <guid isPermaLink="false">50c172c6-74a0-4090-a626-6f6df1236e26</guid>
      <pubDate>Wed, 30 Sep 2026 21:59:39 GMT</pubDate>
      <category>patch_now</category>
      <category>corrected</category>
      <source url="https://api.github.com/advisories?ecosystem=pip&amp;severity=high&amp;per_page=100">GitHub advisories — pip</source>
      <description>[CORRECTED] CVE-2026-102273 (GHSA-w2cx-738m-mc7w) published 29 September 2026, rated high severity. PyJWT 2.13.0 accepts public JWK containers as HMAC secrets. Only version 2.13.0 is listed as vulnerable. Patched in 2.14.0.

Why it matters: An attacker who knows the public key can forge HS256/HS384/HS512 tokens, but only in applications that allow both HS* and asymmetric algorithms in one verification path and pass raw public JWK or JWKS JSON as the key. Applications that keep symmetric and asymmetric verification separate are not affected.

What to do: Check your dependency trees for PyJWT 2.13.0. If found, update to 2.14.0 this cycle. If any verification path accepts both HS* and asymmetric algorithms and passes JWK or JWKS data as the key, review those call sites first.

Original text (as first published, superseded):

HEADLINE: PyJWT accepts JWK public keys as HMAC secrets

WHAT CHANGED: CVE-2026-102273 published 29 September 2026. PyJWT before 2.14.0 accepts public JWK containers as HMAC secrets, rated high severity. Patched in 2.14.0.

WHY IT MATTERS: An attacker who controls the JWK input can forge tokens by supplying a public key where the application expects an HMAC secret, bypassing signature verification entirely. Any JWT validation using HMAC with externally-supplied keys is compromised.

WHAT TO DO: Grep your dependency trees for PyJWT &lt;2.14.0. If found, update to 2.14.0 this cycle. If you pass JWK data into JWT verification, audit those call sites for untrusted input before deploying the patch.</description>
    </item>
  </channel>
</rss>
