Recherche « vercel ai »

Vercel deepsec

Vercel deepsec from 0.1.0 before 1.1.17 contains a link following vulnerability in extractTarballLocally that allows sandboxed code to write files outside the sandbox by planting links in result tarballs. Attackers controlling sandbox code, such as a prompt-injected coding agent, can plant symbolic or hard links in the host data directory to write arbitrary files on the orchestrator host.

vercel satori · vercel next

Satori is a library to convert HTML and CSS to SVG. Starting in version 0.0.27 and prior to version 0.33.5, Satori does not properly escape certain values before including them in generated SVG output. This can allow crafted values to be interpreted as SVG markup. The impact depends on how the generated SVG is consumed. Version 0.33.5 contains a patch. No complete workaround exists besides upgrading. Applications that cannot immediately upgrade should not render attacker-controlled content with Satori.

payloadcms payload

The @payloadcms/storage-vercel-blob storage adapter for Payload contains an improper access control vulnerability that allows authenticated users to bypass collection-level permissions by accessing the client-upload route directly. Attackers can upload files through the client-upload endpoint without possessing the required collection access permissions, circumventing the intended access control enforcement.

baptisteArno typebot.io

TypeBot is a chatbot builder tool. Prior to 3.18.0, any authenticated non-guest workspace member can remove another workspace's public custom domain and make typebots on that domain unavailable. The custom-domain delete handler in handleDeleteCustomDomain.ts authorizes a caller against a client-supplied workspaceId but sends the client-supplied domain name to the shared Vercel project before verifying that the domain belongs to that workspace. This issue is fixed in version 3.18.0.

pydantic-ai · pydantic-ai-slim · Pydantic pydantic ai

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. In versions 1.88.0 up to but not including 1.107.1 and 2.0.0b1 up to but not including 2.5.0, the UI adapters (AG-UI via Agent.to_ag_ui()/AGUIAdapter, and Vercel AI via VercelAIAdapter) use sanitize_messages to strip unresolved ("dangling") client-submitted tool calls from untrusted message history before it reaches the agent, a defense-in-depth default that prevents the agent from executing tool calls the model never emitted. However, the strip anchored to a message index computed before sanitization ran, so when a trailing client message sanitized to empty and was dropped (for example a client system message under the default manage_system_prompt='server'), a preceding assistant response carrying an unresolved tool call became the new tail and was dispatched without inspection. As a result, a remote client could cause a registered, non-approval server tool to run with client-supplied arguments rather than arguments the model produced. The impact is bounded by what the affected tools do and is most significant for applications that gate tool execution in a model-request hook (before_model_request / after_model_request), since a forged call skips the model turn and bypasses that guardrail; approval-gated tools (requires_approval=True) are not auto-executed by this path. This issue has been fixed in versions 1.107.1 and 2.5.0.

pydantic-ai · pydantic-ai-slim · Pydantic pydantic ai

Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.65.0 through 1.105.0, and 2.0.0b1 through 2.0.0b5, a client that submits message history to a Pydantic AI UI adapter (such as the Vercel AI adapter) can reference arbitrary files in the application's model-provider or cloud-storage account. While file URL parts are validated against a scheme allowlist, UploadedFile references — which point to a file by provider file ID or cloud-storage URI (e.g. s3://…, gs://…) — were forwarded without validation. Because the provider resolves an UploadedFile using the server-side identity (IAM role, service account, or provider API key) rather than the client's, an attacker can craft message history to make the server read objects from its own account or other tenants, given a referenceable identifier. Exploitation requires a valid file identifier, which is not always unguessable depending on how the application names objects. This issue has been fixed in versions 1.106.0 and 2.0.0b6.

pydantic-ai · pydantic-ai-slim · Pydantic pydantic ai

Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.56.0 through 1.98.0, when an application opts a URL into force_download='allow-local' (disabling the default block on private/internal IPs), the cloud-metadata blocklist could be bypassed by encoding the metadata IP in an IPv6 transition form (IPv4-mapped IPv6, 6to4, or NAT64), exposing cloud IAM short-term credentials on dual-stack or translated networks. This is an incomplete fix of GHSA-2jrp-274c-jhv3 / CVE-2026-25580, whose remediation did not hold for IPv6-encoded forms of the metadata IPs. An application is affected only if it explicitly opts a FileUrl (ImageUrl, AudioUrl, VideoUrl, DocumentUrl) into force_download='allow-local' on a URL influenced by untrusted input; it is not affected when using bundled integrations to ingest user input (Agent.to_web / clai web, VercelAIAdapter, AGUIAdapter / Agent.to_ag_ui), since they do not propagate force_download from external data, nor when downloading only from developer-controlled URLs. This issue has been fixed in version 1.99.0.

vercel next.js

Next.js is a React framework for building full-stack web applications. In versions 14.1.1 through 15.5.20 and 16.0.0 through 16.2.10, when a Server Action forwards or redirects a request, an attacker can cause the server to send that outbound request to a malicious host (Server-Side Request Forgery). This requires the attacker's request to control Host-associated headers. In some configurations, it's also possible to obtain internal values that weaken middleware/proxy authorization. Applications that use Server Actions are affected when the incoming host header is not fixed to a trusted value. This typically occurs on custom servers, or on deployments not behind a proxy that pins the host. Managed hosting pins the host upstream and is not affected; next start and standalone output do the same from version 14.2 onward. This issue has been fixed in versions 15.5.21 and 16.2.11.

vercel next.js

Next.js is a React framework for building full-stack web applications. In versions 15.5.0 through 15.5.20 and 16.0.0 through 16.2.10, when self-hosting Next.js with the default image loader, the Image Optimization API can optimize remotely hosted images if configured (not enabled by default). If those images contain malicious content, they can cause CPU exhaustion in /_next/image endpoints.Only config.images.remotePatterns is affected, and just the patterns in that array, whereas config.images.unoptimized: true, config.images.loader: 'custom', and Vercel are not impacted. This issue has been fixed in versions 15.5.21 and 16.2.11.

vercel next.js

Next.js is a React framework for building full-stack web applications. In versions 12.0.0 through 15.5.20 and 16.0.0 through 16.2.10, Next.js applications using App Router, Server Actions (use server) or use cache endpoints can be disclosed bypassing any authentication on the pages where these endpoints are usually used. Server Action IDs can be disclosed to unauthenticated users via publicly served client artifacts (for example, static chunks containing action references). Affected users are applications using App Router and Server Actions. By itself, this disclosure is typically a recon/enumeration primitive; however, it can increase risk when combined with other weaknesses. This issue has been fixed in versions 15.5.21 and 16.2.11.

vercel @ai-sdk/harness-opencode

The `@ai-sdk/harness-opencode` tool connects HarnessAgent to OpenCode through a sandboxed bridge. Prior to version 1.0.28, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (`host-tool-mcp.mjs`). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.28 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the OpenCode harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code.

vercel @ai-sdk/harness-codex

The `@ai-sdk/harness-opencode` tool is an HarnessV1 adapter backed by @openai/codex-sdk, which drives the codex command line interface. Prior to version 1.0.29, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (the Codex CLI shim). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools, including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.29 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the Codex harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code.

vercel ai

A vulnerability was determined in vercel ai up to 3.0.97. The impacted element is the function createJsonResponseHandler/createJsonErrorResponseHandler of the file packages/provider-utils/src/response-handler.ts of the component provider-utils. This manipulation causes resource consumption. The attack may be initiated remotely. The exploit has been publicly disclosed and may be utilized. The vendor was contacted early about this disclosure but did not respond in any way.

vercel turborepo

Turborepo is a high-performance build system for JavaScript and TypeScript codebases. Prior to 2.9.14, Turborepo's self-hosted login and SSO browser flows did not validate a CSRF state value on the localhost callback. While the CLI was waiting for authentication, a malicious web page could send a request to the local callback server with an attacker-controlled token. If accepted before the legitimate callback, the CLI could complete login with the wrong credentials. This affects users authenticating the turbo CLI against self-hosted remote cache/auth endpoints. Vercel-hosted login flows using device authorization are not affected. This vulnerability is fixed in 2.9.14.

vercel next.js · Red Hat Streams for Apache Kafka 2.9.4 · Red Hat Trusted Artifact Signer 1.3 · Red Hat Trusted Artifact Signer 1.4

Next.js is a React framework for building full-stack web applications. From 10.0.0 to before 15.5.16 and 16.2.5, when self-hosting Next.js with the default image loader, the Image Optimization API fetches local images entirely into memory without enforcing a maximum size limit. An attacker could cause out-of-memory conditions by requesting large local assets from the /_next/image endpoint that match the images.localPatterns configuration (by default, all patterns are allowed). This vulnerability is fixed in 15.5.16 and 16.2.5.

vercel · Vercel vercel

Vercel’s AI Cloud is a unified platform for building modern applications. From 50.16.0 to 52.0.0, hen the Vercel CLI runs in non-interactive mode (--non-interactive or auto-detected AI agent), commands that cannot complete autonomously emit JSON payloads with suggested follow-up commands. If the user authenticated via --token or -t on the command line, the token value is included verbatim in those suggestions. The plaintext token may be captured in CI/CD logs, agent transcripts, or other automation output. This vulnerability is fixed in 52.0.1.