Headline 1 of 5
Latest news
SecurityMistune fixes CPU-exhaustion flaw in versions before 3.3.4A security advisory reports a denial-of-service flaw in Mistune versions before 3.3.4, with Mistune 3.3.4 as the first fixed version. The flaw lets an ineffective limit on parenthesis scanning drive CPU use up quadratically as a Markdown document grows. Rendering untrusted Markdown can consume substantial CPU; administrators should check their installations and upgrade to Mistune 3.3.4 or later.
SecurityMCP Python SDK fixes OAuth flaw that let servers redirect client credentialsMaintainers of the Model Context Protocol Python SDK published a security advisory for a flaw in its OAuth client support. The advisory covers releases from 2.0.0a1 up to, but not including, 2.2.0, and from 1.9.1 up to, but not including, 1.30.0; it identifies 2.2.0 and 1.30.0 as patched versions. A malicious or compromised MCP server could redirect sensitive authorization material, and users should upgrade to the patched version for their release line.
SecurityGitHub warns of containerd crash risk in three release linesGitHub published a containerd security advisory on 2026-11-19. The advisory lists affected releases as 2.2.0 to before 2.2.10, 2.3.0 to before 2.3.7, and 2.4.0 to before 2.4.2; fixed versions are 2.2.10, 2.3.7, and 2.4.2. Affected versions may crash during image processing and disrupt container operations on a worker node; the advisory recommends upgrading and limiting images to trusted sources, including trusted registries.
Securityeml_parser fixes email-parsing flaw that could stall workerseml_parser maintainers resolved a denial-of-service vulnerability in Received-header parsing affecting versions through 3.0.3. The flaw let regex backtracking consume excessive processor time when applications parsed untrusted email, while the fix reduces the parsing work to O(n). The issue affected service availability only; applications should upgrade to 3.0.4 or later and monitor parser resource use and queue backlogs.
SecurityCodeWhale warns automation defaults can approve agent actions without reviewCodeWhale reported a security flaw in its automation system affecting both packages through version 0.8.67. The advisory says omitted auto-approval settings default to true, unlike the task-path safeguard added for CVE-2026-45374. Recurring agent runs can silently approve tool actions without human review; the advisory lists 0.9.0 as the patched version for both packages.
SecurityCodeWhale advisory flags approval bypass in interactive Agent profileA GitHub security advisory reports that CodeWhale’s default interactive Agent profile can run external programs without asking for approval. The flaw affects the npm package codewhale and Rust package codewhale-tui through version 0.8.67; the advisory lists 0.9.0 as patched for both. The issue can allow arbitrary external program execution on a developer’s machine without an approval prompt; users should update to version 0.9.0.
SecurityUnauthenticated access to k8s-mcp-server could expose Kubernetes privilegesA security advisory published on November 19 reports that k8s-mcp-server can allow unauthenticated clients with network access to its listening port to use Kubernetes tools. The advisory says the default SSE transport lacks authentication, a session token and an Origin check for tool calls, leaving the service’s configured cluster privileges exposed. Versions earlier than 1.1.0 are affected; administrators should upgrade to 1.1.0 and restrict network access to the service.
SecurityGuard Core warns of pipguard-core flaw affecting versions 3.5.0 to 3.12.0Guard Core warned that pipguard-core versions 3.5.0 through 3.12.0 are vulnerable to a denial of service. The advisory says built-in detection patterns run without the timeout protection used for custom patterns, a regression from 3.4.0. Administrators should upgrade to the patched version 3.13.0.
SecurityGuard Core flaw disables two detectors in default modeA GitHub security advisory reports a detection bypass in Guard Core’s default mode. The advisory says whitespace preprocessing disables two built-in detectors for HTTP splitting and command injection. The affected pipguard-core package is patched in version 3.13.0.
SecurityGuard Core reports vulnerability in guard-core packageThe Guard Core advisory reports a vulnerability in the guard-core Python package affecting versions = 1.0.0, < 3.13.0. It says an unvalidated X-Forwarded-For value can be used in Redis keys and pattern matching when the selected proxy-header value is attacker-controllable. The reported risk is to behavioral analytics and the shared Redis instance, with no data exposure reported; version 3.13.0 fixes the issue, and default settings are not affected.
SecurityGitHub advisory flags local code-execution flaw in NLTKGitHub published an advisory on a local code-execution vulnerability in NLTK on November 19, 2026, affecting versions through 3.10.3. The advisory says code could run with the invoking user’s privileges in certain locally configured environments. The advisory recommends upgrading to 3.10.4, the listed patched version.
SecurityNLTK 3.10.0–3.10.3 advisory flags file metadata exposure outside sandboxAn NLTK security advisory reports that StreamBackedCorpusView can expose metadata about files outside its sandbox when applications pass attacker-influenced file IDs directly to its constructor. The advisory says the unguarded metadata check remains possible despite pathsec.ENFORCE and affects 3.10.0–3.10.3, following the content-read fix for CVE-2026-63312. The advisory identifies 3.10.4 as the patched version; file contents remain protected, and exposure depends on how an application constructs the view.
SecurityGitHub advisory flags affected NLTK versionsGitHub published a medium-severity security advisory for NLTK on November 19, 2026, identifying GHSA-cj8f-5fp3-6m88. The advisory lists NLTK versions 3.10.3 and earlier as affected and version 3.10.4 as patched. Users of affected NLTK versions should upgrade to 3.10.4 or later to address the potential for CPU exhaustion.
SecurityGitHub advisory flags NLTK versions through 3.10.3GitHub published a security advisory for NLTK on November 19, 2026. The advisory lists NLTK versions through 3.10.3 as affected and identifies 3.10.4 as patched, while containing conflicting fix-status information. The reported impact is excessive CPU use that could reduce a processing service’s availability; the listed remedy is to upgrade to 3.10.4 or later.
SecurityNLTK advisory flags availability risk through version 3.10.3A GitHub security advisory reports a denial-of-service risk affecting NLTK versions through 3.10.3. The reported availability impact involves malformed WordNet corpus data causing recursion failure during traversal or an unhandled worker process. The advisory text does not establish a confirmed fixed version or consistent official remediation.
SecurityNLTK regex flaw can tie up CPU before its timeout appliesA security advisory reports a compile-time regular-expression denial-of-service flaw in NLTK versions 3.10.0 through 3.10.3. Untrusted patterns can consume excessive CPU because NLTK’s timeout covers matching but not the regular-expression engine’s compilation work. The advisory lists 3.10.4 as patched, but that version had not yet been published; administrators should prevent untrusted pattern input to affected functions and upgrade when a fixed release is available.
SecurityGitHub advises NLTK users to upgrade to version 3.10.4GitHub published a medium-severity security advisory for NLTK, identified as CVE-2026-129228. The advisory lists 3.10.3 and earlier versions as affected and 3.10.4 as patched. Users of affected NLTK versions should upgrade to version 3.10.4 or later.
SecurityGitHub advises NLTK users to upgrade to 3.10.4 or laterGitHub published a medium-severity advisory for the NLTK package on November 19, 2026, identifying CVE-2026-129227. The advisory lists NLTK versions 3.10.3 and earlier as affected and 3.10.4 as patched. Readers should check their NLTK installations and upgrade to 3.10.4 or later.





