Intelligence
Explorer.

msgpack5 is a msgpack v5 implementation for node.js and the browser. Prior to 6.1.0, the streaming decoder recursively invokes itself for each complete MessagePack value remaining in a chunk. A remote peer can send one chunk containing many small valid values, causing recursion proportional to the value count, exhausting the JavaScript call stack, and interrupting the process or stream. This issue is fixed in version 6.1.0.

msgpack5 is a msgpack v5 implementation for node.js and the browser. Prior to 6.1.0, the streaming decoder treats the reserved MessagePack byte 0xc1 as incomplete input instead of invalid input. When 0xc1 begins a stream, subsequent data remains buffered while the decoder waits for bytes that cannot make the value valid, allowing a remote peer to exhaust memory. This issue is fixed in version 6.1.0.

msgpack5 is a msgpack v5 implementation for node.js and the browser. Prior to 6.1.0, the array and map decoding paths have no nesting-depth limit, allowing an attacker who can provide MessagePack input to submit deeply nested containers that exhaust the JavaScript call stack and interrupt a process, worker, or request handler. This issue is fixed in version 6.1.0.

msgpack5 is a msgpack v5 implementation for node.js and the browser. Prior to 6.1.0, the streaming decoder reparses an incomplete array or map from the beginning whenever another chunk arrives. A remote peer can split one valid MessagePack container across many small chunks, causing completed elements to be decoded repeatedly, producing quadratic CPU use and blocking the event loop. This issue is fixed in version 6.1.0.

msgpack5 is a msgpack v5 implementation for node.js and the browser. Prior to 6.1.0, decoding a negative signed 64-bit integer modifies the corresponding bytes in the caller-provided input buffer while computing the value. Applications that retain or reuse encoded input for integrity checks, logging, or later processing can observe silently corrupted data, while positive integers and other MessagePack value types are unaffected. This issue is fixed in version 6.1.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.34.0 until 1.107.4 and 2.28.0, the Agent.to_web() and clai web development chat endpoint has missing request content-type validation. A website visited by a developer can submit a browser-compatible request to a loopback-hosted chat server, causing the served agent to run and execute tools with the privileges and credentials of the local process; client-relayed approval decisions also leave requires_approval=True tools exposed. Binding to localhost does not prevent a browser page from reaching the loopback address. This issue is fixed in versions 1.107.4 and 2.28.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.2 and 2.24.0, web_fetch_tool, the WebFetch local fallback, and remote FileUrl media downloads buffer the complete HTTP response body before enforcing content-size controls. An attacker-influenced URL can stream an arbitrarily large response that exhausts process memory and crashes the worker; affected media types include ImageUrl, DocumentUrl, VideoUrl, and AudioUrl. SSRF protections remain effective, and the impact is limited to availability. This issue is fixed in versions 1.107.2 and 2.24.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 0.3.4 until 1.107.4 and 2.27.1, OpenTelemetry instrumentation configured with InstrumentationSettings(include_content=False) can export retry prompts outside tool calls in gen_ai.input.messages and pydantic_ai.all_messages. Agents using NativeOutput, PromptedOutput, or output validators on text output can therefore disclose validation feedback, including invalid model values quoted by that feedback, to readers of the telemetry backend. Tool-call retries and deployments that do not use include_content=False are not affected by this specific path. This issue is fixed in versions 1.107.4 and 2.27.1.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.34.0 until 2.30.0, the Agent.to_web() and clai web development chat server does not validate the Host header, allowing a website visited by a developer to use DNS rebinding to reach a loopback-hosted agent as a same-origin service. The hostile page can read the served UI and submit chat requests that execute agent tools with the local process's privileges and credentials, causing data disclosure or unwanted side effects. Binding to localhost, Origin checks, and CSRF tokens do not prevent the same-origin DNS rebinding path. This issue is fixed in versions 1.107.5 and 2.30.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 0.3.4 until 1.107.6 and 2.44.0, OpenTelemetry instrumentation configured with InstrumentationSettings(include_content=False) can still export sensitive agent content through exception.message and exception.stacktrace events, error status descriptions, and model_request_parameters containing instructions or the prompted_output_template. The exposed data is available to readers of the configured telemetry backend and can include tool feedback, provider error bodies, runtime instructions, and structured-output templates even though message attributes are redacted. This issue does not grant new access to agent data, and deployments that do not use include_content=False are not affected by the setting bypass. This issue is fixed in versions 1.107.6 and 2.44.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.6 and 2.44.0, the local web_fetch_tool and the WebFetch local fallback process server-controlled responses with quadratic title extraction, whitespace normalization, and ordered-list numbering. An attacker-controlled page of modest size can therefore block the event loop for an extended period, stalling other agent runs and requests, while unsupported codecs or excessive HTML or JSON nesting can abort an individual run. This issue is fixed in versions 1.107.6 and 2.44.0.

Deserialization of Untrusted Data vulnerability in MainWP MainWP Child mainwp-child allows Object Injection.This issue affects MainWP Child: from n/a through 6.2.1.

Obsidian Desktop before 1.14.0 contains a filter bypass vulnerability in the bundled MathJax 3.2.2 Safe component that allows attackers to execute arbitrary code by embedding a crafted \href value with a TAB byte in the URL scheme, causing filterURL to produce an empty protocol that bypasses the configured safeProtocols restrictions. Attackers can craft a note containing a malicious MathJax formula that renders as a javascript: URL anchor, which when clicked by the victim in Live Preview executes in the Node-integration-enabled vault renderer via require('child_process'), achieving arbitrary operating system command execution as the desktop user.

Obsidian Desktop before 1.14.0 contains a remote code execution vulnerability that allows attackers to craft malicious Markdown notes exploiting insufficient sanitization of the data-background-iframe attribute, which bypasses DOMPurify and is processed by the bundled Reveal.js 4.3.1 within the Slides core plugin, allowing a javascript: URL to execute in the resulting background iframe. Because Node integration is enabled and context isolation is disabled in Obsidian's vault renderer, the injected script can call parent.require() to access Node APIs such as fs and child_process, enabling arbitrary operating system command execution when the victim opens the note and manually starts the presentation.

An issue in the ConfirmNameConstraints() function (wolfcrypt/src/asn.c) of wolfSSL v5.9.1 and v5.9.2 allows attackers to cause a Denial of Service (DoS) via providing crafted Certificate Authority certificates, leading to valid certificates without SAN to be incorrectly rejected by wolfSSL-based TLS clients.

HashiCorp go-getter versions before 1.8.10 and go-getter/v2 versions before 2.2.5 are vulnerable to path traversal during S3 and GCS directory downloads, which may allow files to be written outside the requested destination. This vulnerability (CVE-2026-19585) is fixed in go-getter 1.8.10 and go-getter/v2 2.2.5.

FFmpeg through 9.0.2 contains a stack exhaustion vulnerability in av_encryption_init_info_free() in libavutil/encryption_info.c, which recursively frees AVEncryptionInitInfo linked lists built by the MOV demuxer's mov_read_pssh(). Attackers can supply a crafted MP4 file with tens of thousands of small pssh boxes to exhaust the stack and crash the process, while also causing quadratic CPU consumption.

FFmpeg through 9.0.2 contains a denial of service vulnerability in the DASH demuxer that allows attackers to trigger an infinite loop by supplying an empty SegmentTemplate media URL. Attackers can craft an .mpd manifest declaring SegmentTemplate media="" so get_current_fragment() calls av_strireplace() with an empty search string, consuming CPU indefinitely.

FFmpeg through 9.0.2 contains an uninitialized memory disclosure vulnerability in av_dynamic_hdr_plus_to_t35() that leaves up to three payload bytes uninitialized when tone_mapping_flag is 0. Attackers can supply crafted Matroska T.35 BlockAdditional or HEVC/AV1 SEI metadata so that remuxing or transcoding writes leaked process memory into output files.

FFmpeg through 9.0.2 contains a missing host key verification vulnerability in the libssh-based sftp protocol handler that allows network attackers to impersonate SFTP servers. Attackers performing man-in-the-middle, DNS, or ARP spoofing can capture passwords supplied in sftp URLs, serve forged media, or receive uploaded output.

FFmpeg before 8.1.3 and 9.x before 9.0.2 contains an improper certificate validation vulnerability in tls_open() of libavformat/tls_mbedtls.c, which skips hostname checks for IP-address hosts. Network attackers can intercept https, rtmps, or tls connections to IP-literal URLs with any trusted CA-issued certificate to read and tamper with streams.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.56.0 until 1.107.6 and 2.44.0, applications that opt attacker-influenced URLs into local network access through FileUrl with force_download='allow-local' or web_fetch_tool with allow_local_urls=True can bypass the cloud-metadata blocklist by appending an IPv6 zone identifier to an IPv6 metadata address. IPv6Address equality and hashing include the zone identifier, so the blocklist comparison fails even though the network stack ignores the zone on a non-link-local destination and reaches the metadata service, potentially exposing cloud IAM credentials. The opt-in settings are disabled by default, and the issue requires an IPv6-enabled environment. This issue is fixed in versions 1.107.6 and 2.44.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.6 and 2.44.0, the local web_fetch_tool and the WebFetch local fallback compare blocked_domains entries with a URL hostname before both values are normalized to the form used by getaddrinfo. An attacker-influenced model can use an equivalent IDNA spelling, non-ASCII label separator, case variation, or trailing root label that resolves to a blocked host but does not match the configured string, causing the application to fetch that host with its own privileges. allowed_domains fails closed for unmatched spellings, and private-IP and cloud-metadata protections remain effective. This issue is fixed in versions 1.107.6 and 2.44.0.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.7 and 2.52.0, the local web_fetch_tool and the WebFetch local fallback can consume excessive CPU and memory during HTML-to-Markdown conversion of attacker-controlled HTML containing deeply nested block elements. Conversion repeatedly reprocesses accumulated text and can greatly expand intermediate output before the returned-content limit is applied, allowing a model-directed fetch to delay other work in the process. Provider-native web fetching is not affected. This issue is fixed in versions 1.107.7 and 2.52.0.