| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, ToolsMCPServer.run_sse and launch_tools_mcp_server bind to 0.0.0.0 and create /sse and /messages/ routes without invoking the available SecurityConfig authentication, origin-validation, or DNS-rebinding controls. Any reachable client can list and invoke registered tools, and a browser can target a local instance through DNS rebinding, with impact determined by the registered file, shell, and code-execution tools. This vulnerability is fixed in praisonaiagents 1.6.59. |
| On affected versions of Zscaler client connector, a pre-installed peer app can tear down the Zscaler tunnel, force user logout, and toggle packet capture. |
| PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, execute_code sandbox mode permits runtime assembly of blocklisted dunder names and allows str.format or str.format_map to resolve dotted fields through C-level attribute access that bypasses _safe_getattr. This exposes class, qualified-name, base-class, globals, and object-dictionary attributes to prompt-influenced code when approval is automatically granted, producing a high-impact read primitive without establishing a complete in-process execution chain. This issue is fixed in praisonaiagents 1.6.59. |
| gRPC-Go is the Go language implementation of gRPC. Prior to 1.82.2 and 1.83.2, servers created with xds.NewGRPCServer() allow internal/transport/http2_server.go to accept an RPC containing neither the :authority header nor the Host header, while RouteAndProcess in internal/xds/server/routing.go assumes that an authority value exists and indexes the empty slice. A remote client that can complete transport connection establishment can trigger an index-out-of-bounds panic that is not recovered by the per-RPC goroutine and terminates the entire server process. In insecure or ordinary TLS deployments the request can be unauthenticated, while strict mTLS or ALTS deployments require valid transport credentials before the malformed RPC can reach the interceptor. This issue is fixed in versions 1.82.2 and 1.83.2. |
| Pimcore is an Open Source Data & Experience Management Platform. Prior to 2026.1.5, an authenticated user with the objects permission can submit a malicious ClassDefinition UID because the name and ID validation expressions in models/DataObject/ClassDefinition.php validate only the beginning of each value. When a data object of that class containing a Block field is loaded, Block::load in models/DataObject/ClassDefinition/Data/Block.php incorporates the stored class ID into an unquoted object table identifier, allowing the UID to supply SQL syntax. The resulting query can read or modify arbitrary Pimcore database tables, including disclosure of password hashes, and the flaw represents an incomplete validation hardening because earlier work added a start anchor without enforcing the end of the identifier. This issue is fixed in version 2026.1.5. |
| XWiki Platform is a generic wiki platform. Prior to 17.10.5 and 18.2.0, the /skin/ action in com.xpn.xwiki.web.SkinAction can resolve double-encoded parent-directory segments outside the intended skin or web-application resource prefix when Jetty 12 or later decodes the request path. The affected lookup is replaced with Environment.getResourceAsStream(String, String), which constrains a resource to its expected prefix. An unauthenticated remote attacker can use the vulnerable behavior to read arbitrary resources permitted to the Jetty process, including WEB-INF/xwiki.cfg and, depending on deployment depth and operating-system permissions, host files. Tomcat and Jetty versions before 12 do not appear affected. This issue is fixed in versions 17.10.5 and 18.2.0. |
| SpiceDB is an open source database system for creating and managing security-critical application permissions. From 1.34.0 until 1.54.0, SpiceDB can return PERMISSIONSHIP_HAS_PERMISSION instead of PERMISSIONSHIP_CONDITIONAL_PERMISSION or PERMISSIONSHIP_NO_PERMISSION because checkRequestToKey() and checkRequestToKeyWithCanonical() in internal/dispatch/keys/computed.go omit CheckHints when constructing dispatch Check cache keys. The incorrect result requires a permission combining relations with intersection or exclusion, a subject reachable through caveated and non-caveated branches, LookupResources with a context parameter running concurrently with CheckPermission or CheckBulkPermissions for the same resource and subject, and an enabled dispatch result cache. Under these conditions, a result computed for one hint set can poison the cache entry used by a semantically different authorization check, allowing permission without satisfying the caveat. This issue is fixed in version 1.54.0. |
| MikroTik RouterOS before 7.23.4 (long-term) and 7.24.2 (stable) contains a stack-based buffer overflow vulnerability in the mtget binary's TFTP RRQ builder function that allows authenticated users to crash the mtget worker process by supplying a URL path of 507 bytes or more to the /tool fetch command; the first write outside the 528-byte buffer occurs at 505 bytes. Attackers can trigger the overflow by issuing a fetch command with a crafted tftp:// URL path, which causes an unbounded rep movsb instruction to overwrite saved registers at a deterministic offset, crashing the process without requiring a reachable TFTP server or elevated privileges beyond read-only group membership. |
| In CPython 3.13 and earlier, the tarfile module's data and tar extraction filters are vulnerable to crafted archives containing a hard link to a symbolic link. Such archives may cause extraction to modify the permissions or modification time of a file outside the destination directory, or expose the contents of that file within the extracted tree. |
| Casdoor through 4.4.0 fails to properly mask the instance-wide built-in certificate private key in /api/get-certs and /api/get-cert endpoints, allowing organization administrators to retrieve it. Attackers can use the exposed private key to forge JWT tokens for any user in any organization, including global administrators. |
| Crawlab through 0.6.3 uses a hard-coded HMAC-SHA256 secret for JWT token signing that cannot be overridden via configuration or environment variables. Unauthenticated attackers can forge valid administrator tokens to access administrative APIs and execute code on worker nodes. |
| DeepWiki-Open through commit d92819a contains an arbitrary file read vulnerability in the unauthenticated /ws/chat WebSocket endpoint that accepts repo_url as a filesystem path with no containment. Attackers can supply arbitrary directory paths to read all files with supported extensions including Python, JavaScript, YAML, and JSON files containing hardcoded secrets and credentials. |
| Docs through 5.6.1 contains a server-side request forgery vulnerability in the cors-proxy endpoint that allows anonymous attackers to make outbound requests by providing a public document UUID. Attackers can exploit DNS time-of-check-time-of-use race conditions and shared address space bypasses to access internal network resources and exfiltrate image content. |
| Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.17, 0.22.3, and 0.23.2, the tract-onnx crate passes the attacker-controlled external_data location from an ONNX model through onnx/src/tensor.rs get_external_resources and joins the value to the model directory without rejecting absolute paths or parent directory components. Loading an untrusted model through model_for_path can therefore make onnx/src/data_resolver.rs MmapDataResolver open an arbitrary local file and place the file contents into model tensors or inference output. Attacker-controlled offset and length fields can also select an out-of-range mapping slice and cause a denial of service, but the flaw does not write files or execute code. This issue is fixed in versions 0.21.17, 0.22.3, and 0.23.2. |
| MikroTik RouterOS before 7.24.2 contains a path traversal vulnerability in the container package OCI/tar image extraction that allows attackers to write files outside the container root by supplying a crafted container image with symlinks pointing to arbitrary paths. Attackers can exploit unsanitized tar member path extraction during container import via /container/add to achieve root-privileged file creation, directory creation, file deletion via overlayfs whiteout, and hardlink creation on the persistent data partition without ever starting the container. The 7.23.x long-term branch does not contain this fix; the container binaries in container-7.23.3.npk and container-7.23.4.npk are byte-identical, and there is no fixed long-term release at the time of publication. |
| The application's role-authorization lookup defaults to granting access when a request handler's name is not present in its table of role requirements, rather than defaulting to deny. Any request handler that is not explicitly registered in this table is reachable by any authenticated user regardless of their assigned role, and any newly added handler is fail-open by default until explicitly added to the table. |
| PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.58, SpiderTools._validate_url calls _host_is_blocked, which checks literal host encodings but does not resolve DNS names before scrape_page, crawl, extract_links, extract_text, or URL-mention fetches connect. An attacker-controlled hostname resolving to a loopback, private, link-local, or cloud-metadata address therefore bypasses the SSRF policy without a rebinding race and can expose internal responses to the agent. This issue is fixed in praisonaiagents 1.6.58. |
| PraisonAI is a multi-agent teams system. Prior to 4.6.59, the unauthenticated Jobs API accepts an absolute or traversing agent_file path in POST /api/v1/runs and passes it to the job executor without a workspace allowlist or boundary check. A remote caller can cause the server to open files accessible to the service account, exposing credentials, keys, environment variables, and other local data. This vulnerability is fixed in 4.6.59. |
| PraisonAI is a multi-agent teams system. Prior to 4.6.59, the CODE_TOOLS wrappers keep _workspace_root as None and pass workspace=None to read_file, search_replace, and apply_diff helpers that enforce path containment only for a truthy workspace. An application that exposes code_read_file, code_search_replace, or code_apply_diff before set_workspace can therefore let prompt-influenced calls read and modify files outside the intended project directory, while explicitly configured workspaces remain effective. This vulnerability is fixed in 4.6.59. |
| PraisonAI is a multi-agent teams system. Prior to 4.6.58, praisonai.jobs.server.create_app mounts praisonai.jobs.router.create_router under /api/v1/runs without authentication or per-job authorization. Network clients can submit attacker-controlled prompts and agent configuration, list and read jobs, stream results, and cancel or delete other jobs, exposing service credentials and connected tool capabilities to unauthorized agent execution. This vulnerability is fixed in 4.6.58. |
