VAITP

sqlparse is a non-validating SQL parser module for Python. Prior to 0.6.0, sqlparse.format(sql, reindent=True) and sqlformat --reindent route attacker-controlled parenthesized tuple lists through ReindentFilter._get_offset() in sqlparse/filters/reindent.py, where _flatten_up_to_token() repeatedly rebuilds and joins the statement prefix. Thousands of offset calculations walk an expanding token tree, producing quadratic CPU consumption for inputs that remain below MAX_GROUPING_TOKENS and causing request delays, reduced throughput, or worker starvation. This issue is fixed in version 0.6.0.

Cleartext storage of sensitive information in the @step and @remote decorator pipeline component in Amazon SageMaker Python SDK before v3.11.0 and v2.256.0 might allow an authenticated remote user to extract the HMAC signing key from SageMaker DescribePipeline API responses and forge valid integrity signatures for specially crafted function payloads, achieving code execution in another user's pipeline execution context within the same AWS account.

An eval() injection vulnerability in the get_list function in modules/meta_parser.py in lllyasviel Fooocus 2.1.854 through 2.5.5 allows remote attackers to execute arbitrary Python code via a crafted styles payload in the EXIF metadata of an uploaded image file.

ModelScope uses PyYAML's unsafe yaml.Loader to parse model configuration files, allowing arbitrary code execution through Python object construction tags. Attackers can craft malicious model repositories with poisoned configuration files that execute code when loaded by users.

A vulnerability in Hugging Face Transformers (versions >= 4.49.0 and <= 5.8.1) allows remote Python files to be written to local disk without user consent when using GenerativePreTrainedModel.load_custom_generate(). The function fetches and caches a remote module file before performing the required trust_remote_code consent check, inverting the security model enforced by other code-loading paths (such as AutoConfig, AutoModel, and AutoTokenizer). As a result, attacker‑controlled Python code from custom_generate/generate.py is copied into the user’s ~/.cache/huggingface/modules directory even if the user declines the trust prompt. Although execution is correctly gated, the file write is not reversible and can persist across sessions. This can lead to persistent, unauthorized files on disk and stale cache collisions where cached attacker code may later be executed during trusted model loads. The issue stems from an unconditional file write in dynamic_module_utils.py prior to any trust verification.

Tornado is a Python web framework and asynchronous networking library. Prior to 6.5.8, Tornado parses application/x-www-form-urlencoded request bodies with urllib.parse.parse_qs in tornado/escape.py without passing max_num_fields. RequestHandler._execute in tornado/web.py parses the body before handler dispatch through HTTPServerRequest._parse_body and parse_body_arguments in tornado/httputil.py, so an unauthenticated request body containing millions of separator-delimited fields can synchronously stall the single-threaded event loop and delay every connection. The body is bounded only by max_buffer_size, which defaults to 104857600 bytes. This issue is fixed in version 6.5.8.

Stored Cross-Site Scripting (XSS) in TaskRunHandler.post() in web/handlers/task.py in QD 20220208 through 20250803. When a task is run via /task/<taskid>/run, the handler renders task log content (logtmp) into the HTML response using Python % string formatting without HTML encoding. logtmp is populated from the exception object or from new_env.variables.__log__, which is attacker-controlled via the template extract_variables mechanism. A low-privileged authenticated attacker can create a crafted HAR template that extracts arbitrary HTML/JavaScript into the __log__ variable via the api://util/unicode endpoint. When a victim triggers the task run, the embedded script executes in the victim browser within the QD application context.

BISHENG before 2.6.0 contains a remote code execution vulnerability in the workflow run_once endpoint that allows authenticated users to execute arbitrary Python code. Attackers can submit crafted Code node definitions to the POST /api/v1/workflow/run_once endpoint, which executes them with exec() without sandboxing, gaining access to filesystem, credentials, and internal network resources.

Piccolo Admin is an admin interface and content management system for Python, built on top of Piccolo. Prior to 1.14.0, piccolo_admin/endpoints.py uses superuser_validators to block PUT, PATCH, DELETE, and POST requests by non-superusers but permits GET requests to configured user and session tables, while piccolo_api/session_auth/tables.py exposes SessionsBase.token because the token column is not secret. In deployments that add the Sessions and User tables to create_admin, a non-superuser administrator can call GET /api/tables/sessions/, obtain another user's live session token, replay it as the Cookie id value to impersonate a superuser, and permanently set superuser to true on the attacker's own row. This issue is fixed in version 1.14.0.

Agno up to and including 2.5.8 is vulnerable to Remote Code Execution (RCE) via prompt injection. The PythonTools and ShellTools components pass unsanitized, LLM-generated arguments directly to execution sinks including exec(), runpy.run_path(), and subprocess.run(). An unauthenticated attacker can exploit this by embedding malicious instructions in content processed by the agent (such as web pages or documents), allowing for arbitrary code and OS command execution on the host server.