PraisonAI UI MCP connect endpoint allows unauthenticated local command execution
PraisonAI is a multi-agent teams system. Prior to 4.6.59, the default UI host applications expose POST /api/mcp/connect without mandatory authentication and accept caller-controlled command and args values that PraisonAIUI passes to StdioMCPClient to start a local process. Because the UI commands bind to 0.0.0.0 by default, a reachable unauthenticated client can execute commands as the UI service account even when the MCP handshake later fails. This vulnerability is fixed in 4.6.59.
PraisonAI: Webhook signature verification skipped (fail-open) when secret unset, allowing forged inbound webhooks (WhatsApp & Linear bots)
PraisonAI is a multi-agent teams system. Prior to 4.6.59, the WhatsApp and Linear bot webhook handlers verify HMAC signatures only when WHATSAPP_APP_SECRET or LINEAR_WEBHOOK_SECRET is configured and otherwise parse and dispatch unsigned request bodies. A remote unauthenticated client that reaches the webhook route can forge messages, comments, or agent-session events, impersonate platform users, influence agent prompts and actions, and disrupt bot processing. This issue is fixed in 4.6.59.
praisonai: recipe serve auth middleware silently disables itself when no secret is set
PraisonAI is a multi-agent teams system. Prior to 4.6.58, recipe serve installs APIKeyAuthMiddleware or JWTAuthMiddleware when an operator selects api-key or JWT authentication, but each middleware forwards requests when PRAISONAI_API_KEY or PRAISONAI_JWT_SECRET and the corresponding recipe value are absent. Unauthenticated clients can then reach recipe execution, input, and output surfaces and may trigger connected tools despite the operator explicitly enabling authentication. This issue is fixed in 4.6.58.
PraisonAI Code agent tools fail open without a workspace boundary
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: Unauthenticated Local File Inclusion via agent_file path in the Jobs API
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: Arbitrary File Read via `@file:` Mention Path Traversal
PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, MentionsParser._process_file_mention accepts file-mention values and falls back from workspace-relative resolution to Path(file_path) without traversal, symlink, or workspace-boundary validation. Prompt input from users, bots, or workflows can therefore read arbitrary files accessible to the process, including credentials, keys, environment files, source code, and system configuration. This issue is fixed in praisonaiagents 1.6.59.
praisonaiagents: SSRF guard validates literal IPs only and never resolves DNS
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 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.
PraisonAI: MCP SSE transport binds 0.0.0.0 with no authentication and no Origin validation; bundled SecurityConfig is never wired in
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.
PraisonAI: IMAP Command Injection via Unsanitized Email Search Parameters
PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, src/praisonai-agents/praisonaiagents/tools/email_tools.py interpolates LLM-controlled from_addr, subject, and query values directly into quoted IMAP SEARCH criteria. Embedded quote, backslash, newline, or null characters can escape the intended criterion and alter IMAP operations when search_emails, reply_email, or archive_email is exposed to an agent with configured email credentials, allowing mailbox data access, modification, deletion, or connection disruption. This issue is fixed in praisonaiagents 1.6.59.
PraisonAI: Unauthenticated Event Injection via SSE `/publish` Endpoint
PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.58, the SSE server in src/praisonai-agents/praisonaiagents/server/server.py does not consult ServerConfig.auth_token before handling /publish, /events, or /info requests. A network client that can reach the server can broadcast arbitrary events to connected clients and obtain server configuration and client-count information. This issue is fixed in praisonaiagents 1.6.58.
PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, SpiderTools.scrape_page validates only the initial URL and lets requests.Session.get follow redirects automatically, so a public-looking URL can redirect to a loopback, private, link-local, or metadata address without revalidation. The redirected response body is returned through scrape_page and its extract_links, crawl, and extract_text callers, allowing disclosure from otherwise blocked services. This issue is fixed in praisonaiagents 1.6.59.
PraisonAI: Unauthenticated RCE via Jobs API + Approval Bypass
PraisonAI is a multi-agent teams system. Prior to praisonai 4.6.59 and praisonaiagents 1.6.59, the unauthenticated POST /api/v1/runs Jobs API accepts attacker-controlled agent_yaml, and the approve field can mark execute_command as YAML-approved before @require_approval checks critical tools. This chain allows a remote caller to cause a configured language model agent to invoke arbitrary operating-system commands without credentials or operator interaction. This vulnerability is fixed in praisonai 4.6.59 and praisonaiagents 1.6.59 as fixed versions.
vLLM before 0.28.0 Remote Code Execution via LlavaOnevision2 processor
vLLM before 0.28.0 contains a remote code execution vulnerability in the LlavaOnevision2 processor loader that ignores the trust_remote_code parameter when loading remote processor classes. Attackers can craft a malicious model with arbitrary code in processing_llava_onevision2.py that executes with vLLM process authority even when trust_remote_code is set to False.
tarfile hardlink fallback ignores custom extraction filter rejection via None
When tarfile extracts a link on a system that doesn't support links, it falls back to extracting a member from the archive. In this case, the filter function is run twice: once for the extracted member, and once with name set to the location of the link. For one of the calls, the return value was ignored. Instead, the member should be skipped if either call returns None.
Headroom vulnerable to Cross-Site WebSocket Hijacking (CSWSH)
Headroom compresses data before the data reaches a large language model. Prior to version 0.35.0, the Headroom WebSocket server does not validate the `Origin` header of incoming client WebSocket requests before forwarding the request to the upstream server, allowing malicious WebSocket clients to perform arbitrary LLM requests without authentication. This can be exploited by a malicious WebSocket client executed in a traditional or headless browser such as lightpanda, if the browser has access to the Headroom proxy and the OpenAI API key is stored in the `OPENAI_API_KEY` environment variable. Version 0.35.0 fixes the issue.
psd-tools: arbitrary file write via smart-object filename
psd-tools is a Python package for working with Adobe Photoshop PSD files. Prior to version 1.17.1, `SmartObject.save()` writes an embedded smart object to a path taken verbatim from the PSD file. Because that name is attacker-controlled and unsanitised, a tool that extracts embedded objects from an untrusted `.psd` can be made to write attacker-chosen bytes to an attacker-chosen path (absolute or `../`-traversing), outside its intended output directory. A secondary issue in `SmartObject.open()` for external-kind smart objects allows the attacker-controlled `fullPath` descriptor to be used as an arbitrary file read path, enabling exfiltration
GridFS data disclosure and deletion via query-operator injection in file IDs in the MongoDB Python Driver
Improper neutralization of special elements in data query logic in the GridFS component of the MongoDB Python Driver can cause a caller-supplied structured file identifier to be interpreted as a query condition rather than as a literal identifier. An authenticated user who can influence the identifier passed by an affected application may obtain stored file content beyond the intended target or cause all GridFS file chunks in the affected bucket to be removed, rendering stored file content unreadable. The affected rename operation may also rename a stored file other than the intended target.
Open WebUI: Users denied by the OAuth role policy can still sign in via token exchange
Open WebUI is an extensible, feature-rich, and user-friendly self-hosted AI platform. From 0.8.0 until 0.11.1, Open WebUI's OAuth token exchange endpoint issues a session for a provider access token without running the OAuth role management that the normal OAuth login callback runs. A user whose provider roles the login callback would refuse, or would demote, could still obtain a working session at their existing role through this endpoint. This issue is fixed in version 0.11.1.
Open WebUI: Users denied by the OAuth domain allowlist or role policy can still sign in via token exchange
Open WebUI is an extensible, feature-rich, and user-friendly self-hosted AI platform. From 0.8.0 until 0.9.0, Open WebUI's OAuth token exchange endpoint issues a session for a provider access token without applying the email domain allowlist that the normal OAuth login callback enforces. An account whose email domain the login callback would refuse could still obtain a working session through this endpoint. This issue is fixed in version 0.9.0.