| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| Improper Neutralization of Special Elements used in an OS Command in the integrate-platform-docs composite GitHub Action of Chainguard Academy (edu) from commit 7375a80caabcc31c33ec90f29687ed78c13d16ff before commit fb0efb2537d326ab18c07d620875b8ed2a4b39f3 allows an actor who controls the project_id or storage_bucket inputs to execute arbitrary shell commands on the GitHub Actions runner, because the inputs are interpolated directly into Bash gcloud storage cp commands in several steps via ${{ inputs.* }} expressions. The only in-repository caller passed repository secrets and ran only on trusted triggers, so no untrusted input was known to reach the vulnerable steps. |
| Dify is an open-source LLM app development platform. Prior to 1.16.0, the PUT /console/api/apps/<app_id>/server endpoint in api/controllers/console/app/mcp_server.py used AppMCPServerController.put() to retrieve an AppMCPServer by the client-supplied server ID without verifying that the server belonged to the requested application and tenant. An authenticated workspace member could therefore change another application's MCP server status and parameters, potentially redirecting data or disabling the service. This issue is fixed in version 1.16.0. |
| vLLM is an inference and serving engine for large language models. Prior to 0.30.0, a caller can use the request-level media_io_kwargs field to select the GLMGA video backend and supply large values for the fps and max_frames options without a strict work ceiling. GLMGA constructs and deduplicates an attacker-sized pre-decode frame-index list, allowing a compact request and tiny valid video to consume disproportionate CPU time and memory in the shared media-loading executor. This issue is fixed in version 0.30.0. |
| vLLM is an inference and serving engine for large language models. Prior to 0.30.0, OpenAI-compatible request models accept a non-empty cache_salt value without enforcing the character and length restrictions required by the IPCCacheServerKey consumer in LMCache-MP. On deployments using the LMCache-MP connector, a salt that contains a forbidden character or exceeds the permitted length can raise an uncaught ValueError during scheduler cache lookup, causing EngineCore to terminate and denying service to all concurrent users. This issue is fixed in version 0.30.0. |
| vLLM is an inference and serving engine for large language models. Prior to 0.30.0, Harmony tool continuations submitted through "POST /v1/responses" requests rebuild the next-turn engine input without preserving the cache_salt value, placing the continuation prefix in the global unsalted cache namespace even when the caller enabled salting. On deployments with prefix caching enabled, which is the default, an authenticated tenant who can reconstruct a victim's low-entropy post-tool history can submit the same continuation and use the cached_tokens_per_turn count to determine whether the prefix was previously processed, defeating the intended tenant isolation of salted prefix caching. This issue is fixed in version 0.30.0. |
| Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.16.0 until 2.131.0, the InputFormat.JSON_DOCLING backend in docling/backend/json/docling_json_backend.py validates serialized DoclingDocument input without rejecting picture image references that contain local paths or file URIs. When the document is enriched or exported with ImageRefMode.EMBEDDED, the DoclingDocument._with_embedded_pictures and ImageRef.pil_image methods can open those references and place readable image bytes in Markdown or HTML output. Disclosure is limited to files Pillow can decode as images, while differing decode behavior can also reveal whether a path exists. Direct untrusted loading through docling-core is outside this Docling fix. This issue is fixed in 2.131.0. |
| Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.45.0 until 2.131.0, METS-GBS format detection in docling/datamodel/document.py and the backend in docling/backend/mets_gbs_backend.py call tarfile.TarFile.getmembers() before enforcing the max_member_count limit, causing the full archive member list to be allocated before the limit can stop processing. A small gzip-compressed tar archive with a very large number of empty members can therefore consume memory proportional to the declared member count, including during format detection before the allowed_formats restriction is applied. This issue is a residual weakness in the member-count protection added for CVE-2026-44018. This issue is fixed in 2.131.0. |
| Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.27.0 until 2.131.0, Docling plugin factories in docling/models/factories/base_factory.py call load_setuptools_entrypoints() before applying the allow_external_plugins setting, so every module registered in the Docling entry-point group is imported even when external plugins are disabled. An installed third-party or compromised package can therefore execute import-time code when Docling starts, while the subsequent namespace filter misleadingly reports that the plugin was not loaded. This issue is fixed in 2.131.0. |
| Langflow is a tool for building and deploying AI-powered agents and workflows. Before Langflow 1.10.3, the MCP stdio transport launched whatever command / args a user put in an MCP server configuration, with no allowlist and (before 1.10.3) wrapped in bash -c "exec {command} ...". Any user able to reach the MCP server settings ("Settings → MCP Servers → Add MCP Server", POST/PATCH /api/v2/mcp/servers/{server_name}) or to build a flow with the MCP Tools component could add a "server" whose command is an arbitrary OS command (touch, rm -rf, a reverse shell, ...). The command runs on the Langflow host as the Langflow process user as soon as Langflow tries to connect to the server (listing servers, loading tools, running the flow) — even when the UI then reports that the stdio server failed to start. With the default LANGFLOW_AUTO_LOGIN=true, GET /api/v1/auto_login hands out a token without credentials, so on an exposed instance running the default configuration this is reachable without an account. AUTO_LOGIN is documented as a development-only setting; with it disabled, any authenticated (non-admin) user can exploit it. This issue is fixed in Langflow 1.10.3, langflow-base 0.10.3, and lfx 1.10.3. |
| Penpot is an open-source design and prototyping platform. Prior to 2.18.0, the delete-share-link RPC retrieves a caller-selected share-link ID and verifies only that the caller can edit the parent file. It does not verify that the caller created the share link or has owner or administrator authority, allowing any file editor who knows a share-link UUID to delete links created by other users and revoke external reviewers' access. This issue is fixed in version 2.18.0. |
| Penpot is an open-source design and prototyping platform. Prior to 2.18.0, the delete-team-member RPC allows a team administrator to remove any member other than themselves but does not protect the team owner. A non-owner administrator can delete the owner's team-profile-rel membership and lock the owner out of the team and its projects, files, fonts, and media. This issue is fixed in version 2.18.0. |
| Ghost is a Node.js content management system. From 5.9.0 until 6.44.1, an input validation issue allowed members to access comments they were not authorized to access. This issue is fixed in version 6.44.1. |
| Ghost is a Node.js content management system. From 1.20.0 until 6.64.0, a vulnerability in how Ghost loads theme translation files allowed an authenticated Administrator to read JSON files outside of the active theme's directory, potentially exposing server configuration secrets. This issue is fixed in version 6.64.0. |
| Ghost is a Node.js content management system. From 4.22.0 until 6.65.0, SVG media thumbnails and SVG images uploaded with a non-SVG file extension were stored without sanitization. This allowed any staff user, including Contributors, to host scripts on the site's domain, possibly resulting in compromise of other staff users' admin sessions. This issue is fixed in version 6.65.0. |
| Ghost is a Node.js content management system. From 4.0.0 until 6.67.0, SVG images included in content imports were stored without sanitization. An attacker who convinced an Administrator to import a crafted file could host scripts on the site's domain, possibly resulting in compromise of staff users' admin sessions. This issue is fixed in version 6.67.0. |
| Plane is an open-source project management tool. Prior to 1.4.0, Plane's signup flow creates a logged-in User row for any submitted email without an out-of-band ownership check, while User.email is unique=True. The authenticated user can call GET /api/users/me/workspaces/invitations/, which returns each WorkspaceMemberInvite whose email matches request.user.email. WorkSpaceMemberInviteSerializer uses fields = "all", exposing the token that protects the invitation join endpoint. An unauthenticated attacker who knows a target's email can register an account using that address, enumerate pending invitations, and accept an invitation as the target, joining a workspace at the invited role. The term pre-auth describes the attacker's initial state: the attacker has no credential before signup, while the enumeration and join requests use the session created by that signup. This issue is fixed in 1.4.0. |
| Plane is an open-source project management tool. Prior to 1.3.0, the ProjectMemberViewSet.partial_update method allows any project member, including a user with the lowest GUEST role, to modify another project member's role. The authorization check prevents assigning a role higher than the requester's role but does not prevent assigning a lower or equal role, allowing a Guest to demote Administrators and Members and deny them project control. This vulnerability is fixed in 1.3.0. |
| Plane is an open-source project management tool. Prior to 1.4.0, BulkEstimatePointEndpoint.destroy resolves an estimate point through a bare primary-key lookup without workspace, project, or estimate scoping. An administrator or member of one workspace can permanently delete an estimate point belonging to another workspace by supplying the target UUID in a URL under the attacker's own workspace. This creates a destructive cross-tenant IDOR. This issue is fixed in 1.4.0. |
| An argument injection flaw was found in CUPS. When email notification is configured, the CUPS scheduler accepts printer subscription requests that supply a mailto notify-recipient-uri. The mailto notifier passes the recipient address to the configured sendmail program without ensuring it cannot be interpreted as command-line options. A remote attacker who can reach the CUPS service could supply a crafted recipient value starting with "-" to influence sendmail behavior. Successful exploitation depends on the installed mail transfer agent and CUPS network exposure, and may lead to execution of attacker-controlled commands with the privileges of the CUPS service user. |
| Plane is an open-source project management tool. Prior to 1.4.0, Plane validates GITEA_HOST only for its URL scheme and does not reject hosts that resolve to private or internal IP addresses. The four outbound requests in the Gitea OAuth flow are derived from this unvalidated host and do not call validate_url(). In addition, avatar_url is taken from the Gitea user's profile, where users can configure external avatar URLs. After an administrator enables Gitea OAuth for a legitimate instance, a Gitea user can set an internal URL as the profile avatar and log in through Gitea, causing Plane to fetch the internal target without validation. This issue is fixed in 1.4.0. |
