| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| In camel-ai camel 0.2.91a1, v0.2.91a2 and v0.2.91a3, TerminalToolkit.shell_exec allows prompt-driven shell command execution without an approval boundary. |
| Dell PowerStore contains a Code Injection vulnerability. An authenticated user with limited privileges could potentially exploit this vulnerability to execute arbitrary code with root privileges. |
| In camel-ai camel 0.2.91a1, v0.2.91a2 and v0.2.91a3, CodeExecutionToolkit can run model-produced Python code through SubprocessInterpreter without an approval boundary. |
| bisheng 2.3.0, 2.4.0, and 2.4.0-beta1 is vulnerable to Code Injection in src/backend/bisheng/api/v1/validate.py. |
| Devika v1.0 is vulnerable to Code Injection via the Runner.run_code function in src/agents/runner/runner.py. |
| OpenPanel js-runtime through 2.3.0 contains a sandbox escape vulnerability in the JavaScript webhook template validator that fails to block computed member access to constructor chains. Attackers with project write access can create webhook templates using computed property notation to access Function constructor and execute arbitrary code in the worker process. |
| OpenPanel before 2.3.0 fails to properly validate chart formula expressions, allowing authenticated project members with read access to execute arbitrary code by recovering the native JavaScript Function constructor through mathjs matrix objects. Attackers can use the recovered constructor to load Node.js built-ins and execute operating system commands with the privileges of the API process, bypassing organization authorization boundaries. |
| A vulnerability was found in mwasikz robo-cafe-rms up to 228c44a02823f04e85db32b7137809a2856148fc. The impacted element is an unknown function of the file frontend/update-account.php of the component Edit Profile Feature. Performing a manipulation of the argument Name/Address/City results in cross site scripting. Remote exploitation of the attack is possible. The exploit has been made public and could be used. This product follows a rolling release approach for continuous delivery, so version details for affected or updated releases are not provided. The vendor was contacted early about this disclosure but did not respond in any way. |
| In agentscope 1.0.18, 1.0.19, and 1.0.19 when the RealtimeAgent session exposes execute_python_code as an available tool, a remote WebSocket user can prompt the agent to call that tool and run Python code in the service environment. In the validated path, RealtimeAgent._acting forwards the model-produced tool call to Toolkit.call_tool_function, which invokes execute_python_code without an additional approval or isolation boundary on that path. |
| djehuty is a research data repository system developed by 4TU.ResearchData. Prior to version 26.3.2, An unauthenticated attacker can inject SPARQL into the search/listing queries through three separate parameters. Because the affected queries are read (SELECT) queries, this does not write to the store, but it allows: Cross-graph data exfiltration — e.g. UNION-ing in triples from graphs the request was never scoped to (drafts/private/internal data held in the RDF store); denial of service — expensive or malformed queries that tie up the SPARQL backend / web workers. No account or user interaction is required. This issue has been patched in version 26.3.2. |
| An issue in Greek Universities Network (GUnet) Open eClass Platform v.3.15 allows a remote attacker to execute arbitrary code via the chat input field in the course module. |
| MindsDB Minds Platform version 26.1.0 and earlier contains an unauthenticated remote code execution vulnerability that allows unauthenticated attackers to execute arbitrary OS commands by submitting crafted prompts to the unprotected POST /api/v1/responses/ endpoint, which reaches the Anton agent's scratchpad tool that calls exec() on attacker-influenced Python source without sandboxing. Attackers can first configure their own LLM API key through the unauthenticated PUT /api/v1/settings/ endpoint, then POST a prompt directing the agent to invoke the scratchpad tool with arbitrary Python code, achieving full OS command execution as the user running the desktop application and enabling access to SSH keys, stored credentials, and environment secrets. |
| In DB-GPT 0.7.5 and 0.8.0, a skill uploaded through the real /api/v1/skills/upload route can later be executed through the real /api/v1/chat/react-agent flow. |
| Devika v1.0 is vulnerable to Code Injection in the Runner.execute function in src/agents/runner/runner.py which allows an attacker to achieve arbitrary code execution by exploiting the direct execution of LLM-generated content. |
| iControl is affected by an insecure Cross-Origin Resource Sharing (CORS) policy vulnerability, which could allow a malicious website to execute cross-origin requests with included credentials, enabling an attacker to access and exfiltrate sensitive data within the context of the victim's active session. |
| Kobako is a Ruby gem that embeds a Wasm-isolated mruby interpreter inside applications, allowing execution of untrusted Ruby scripts (LLM-generated code, user formulas, student submissions, third-party plugins) in-process without giving them access to host memory, files, network, or credentials. From version 0.1.0 to before version 0.9.1, a guest mruby script running inside the Kobako sandbox can execute arbitrary Ruby in the host process, fully escaping the sandbox. This issue has been patched in version 0.9.1. |
| Taskcluster is the task execution framework that supports Mozilla's continuous integration and release processes. Prior to version 100.3.0, Taskcluster is vulnerable to unauthenticated RCE on Taskcluster deployments with an anonymous role that exposes the GraphQL endpoint and parses filter arguments using the sift library. This issue has been patched in version 100.3.0. |
| The The Appointment Booking Plugin – LatePoint | Calendar & Scheduling for WordPress plugin for WordPress is vulnerable to arbitrary shortcode execution in all versions up to, and including, 5.7.0. This is due to the software allowing users to execute an action that does not properly validate a value before running do_shortcode. This makes it possible for unauthenticated attackers to execute arbitrary shortcodes. The payload is planted during the unauthenticated booking flow and triggered when the Customer Cabinet block rendered by render_customer_dashboard() outputs the stored name into the content stream, where WordPress core's do_shortcode filter at priority 11 re-parses and executes it. |
| openssl_encrypt before 1.4.9 executes untrusted third-party plugins with insufficient controls: the plugin signature policy defaulted to WARN, so an unsigned/unverifiable non-built-in plugin was compiled and executed in the host process at import time, before the runtime sandbox is installed. The only default gate was an incomplete, bypassable AST denylist. If a user is induced to load an attacker's plugin, this results in arbitrary code execution with the privileges of the user running openssl_encrypt. Fixed in 1.4.9 by defaulting the signature policy to ENFORCE for non-built-in plugins. |
| Xinference loads models with Hugging Face remote code execution unconditionally enabled, and before version 2.12.0 exposes no setting to disable it. Six loader call sites pass trust_remote_code=True as a literal or as an unconditional default: RerankModel._get_tokenizer in xinference/model/rerank/core.py, SentenceTransformerRerankModel.load in xinference/model/rerank/sentence_transformers/core.py, SentenceTransformerEmbeddingModel.load in xinference/model/embedding/sentence_transformers/core.py, FlagEmbeddingModel.load in xinference/model/embedding/flag/core.py, and two sites in xinference/model/llm/transformers/core.py where PytorchModel._sanitize_model_config and PytorchModel._get_components default the value to True. Because a caller with model launch access can register a model whose type is unknown and supply an arbitrary model path, the server reaches _auto_detect_type and then AutoTokenizer.from_pretrained, which imports and executes Python declared by the model directory's own tokenizer_config.json auto_map, running attacker-supplied code with the privileges of the worker process. Version 2.12.0 gates every site behind allow_trust_remote_code and the XINFERENCE_TRUST_REMOTE_CODE setting, permitting remote code only for bundled built-in models. |
