Intelligence
Explorer.

BotSharp through 5.2.0 contains an authentication bypass vulnerability that allows unauthenticated remote attackers to forge bearer tokens using the hard-coded Jwt:Key in WebStarter appsettings.json. Attackers can sign tokens with the committed HMAC secret and fixed botsharp issuer and audience to impersonate any known user, including administrators, on Authorize-protected API routes.

mcp-go through 1.2.1 contains a denial of service vulnerability in StreamableHTTPServer.ServeHTTP that allows remote unauthenticated attackers to exhaust memory by sending oversized POST bodies. Attackers can send arbitrarily large or many concurrent POST requests, read fully via io.ReadAll before validation, to degrade or OOM-kill the server process.

Predibase LoRAX through 0.12.1 contains a sensitive information exposure vulnerability that writes the caller-supplied api_token from POST /generate request bodies into router logs. Attackers with access to router logs or OTLP trace backends can recover other users' private-adapter tokens recorded through the instrumented GenerateParameters span field.

Hugging Face Text Embeddings Inference through 1.9.4 contains a cleartext logging vulnerability that exposes the configured api_key because the router's Args struct lacks a redact attribute for it. Attackers with access to router logs, container output, or OTLP telemetry can recover the Bearer token and call the protected embedding and rerank endpoints.

UnicomAI Wanwu through 0.6.5 contains an authorization bypass vulnerability that allows authenticated users to mint AppKeys bound to other users' MCP servers via POST /v1/appspace/app/key. Attackers supplying a victim's MCP server UUID with appType mcpserver can open MCP sessions and invoke the server's tools using the victim's upstream authentication.

UnicomAI Wanwu through 0.6.5 contains a missing authorization vulnerability that allows any authenticated enabled user to revoke other users' AppKeys for arbitrary apps via the unpublish endpoint. Attackers can supply a target appId and appType from the exploration marketplace to delete other users' api_key rows across organizations, cutting off MCP and OpenAPI client access.

Wanwu before 0.6.3 contains an insecure direct object reference vulnerability that allows any authenticated enabled user to delete other users' legacy AppKeys by supplying a numeric apiId. Attackers can iterate sequential key IDs against DELETE /v1/appspace/app/key to revoke AppKeys across organizations, breaking MCP and OpenAPI clients until owners issue new keys.

UnicomAI Wanwu before 0.6.3 contains an insecure direct object reference vulnerability that allows authenticated low-privileged users to delete other tenants' agent or RAG applications by supplying their appId. Attackers can send requests to DELETE /v1/appspace/app with guessed sequential assistant IDs to permanently delete victims' applications, workflows, conversations, and associated data.

Deep Chat through 2.5.1 contains a cross-site scripting vulnerability that allows attackers to inject javascript: links because RemarkableConfig.createNew disables Remarkable link validation. Attackers can place crafted Markdown links in AI responses, addMessage content, or loaded history to execute script in the embedding page when victims click them.

phpMyFAQ through 4.1.10 contains a missing authorization vulnerability in the MCP server faq_search tool that allows MCP clients to read restricted FAQs because Search::searchDatabase() never applies user or group permission checks. Attackers connected to the phpmyfaq:mcp:server can issue search queries to retrieve the full question and answer text of active FAQs restricted to specific users or groups.

Company Research Agent through 2.2.0 contains a server-side request forgery vulnerability that allows unauthenticated attackers to trigger outbound requests by injecting unescaped ReportLab paragraph markup into the /generate-pdf endpoint. Attackers can embed inline img elements in report_content to make the server fetch internal or external hosts, leaking image responses in returned PDFs and probing reachability.

thClaws through 0.141.0 contains a link-following vulnerability in the post_inputs handler of the v1 API POST /v1/inputs endpoint that allows writes outside the workspace by following symlinks. Attackers who plant a symlink under the allowed inputs/ prefix can cause later authenticated uploads to write or truncate files outside the workspace with daemon privileges.

hyper-mcp through 0.8.3 contains an improper signature verification vulnerability that allows attackers to load malicious WebAssembly plugins because cosign_verify_args() accepts any signer identity and OIDC issuer by default. Attackers controlling a plugin image reference can sign a malicious image with a free Sigstore keyless certificate to execute plugins with configured host, filesystem, and environment capabilities.

hyper-mcp through 0.8.3 contains a signature verification bypass vulnerability in load_wasm in src/wasm/oci.rs that verifies the Cosign signature of a separately resolved tag rather than the loaded manifest. Attackers controlling registry responses for the tag can serve an unsigned malicious manifest to the loader and a signed one to Cosign, executing unsigned WebAssembly plugins with configured host capabilities.

CoreShop through 2026.2.2 contains a missing authorization vulnerability that allows low-privileged backend users to list permission-restricted resources because ResourceController listAction skips the isGrantedOr403() check. Authenticated Pimcore users lacking resource permissions can request the generated list routes to enumerate payment providers, carriers, price rules, stores, and tax rules including ids, names, and identifiers.

A security flaw has been discovered in zhayujie CowAgent up to 2.1.7. Impacted is an unknown function of the file channel/web/web_channel.py of the component Web Console. The manipulation of the argument session_id results in denial of service. The attack may be performed from remote. The exploit has been released to the public and may be used for attacks. Version 2.1.7 (commit dec28324) only partly mitigates via a 512MB body cap. The vendor was contacted early about this disclosure.

LlamaFarm through 0.0.34 contains an insecure default configuration that binds its unauthenticated FastAPI server to 0.0.0.0 on port 14345, while the lf CLI silently discards HOST overrides. Network-adjacent attackers can call the project and dataset management API to read stored provider API keys, modify projects, trigger ingestion, and irreversibly delete projects.

mistral.rs 0.9.0 through 0.9.4 contains a link following vulnerability in mistralrs-code-exec that allows sandboxed shell code to read and overwrite files outside the sandbox via symlinks. Attackers or prompt-injected agents can name symlinks as outputs or reuse sessions with symlinked input paths to access files with the server process's permissions.

Easy!Appointments through 1.6.0 contains an authorization bypass vulnerability in Booking::register() that allows unauthenticated attackers to modify any appointment by supplying an appointment id without its hash. Attackers can enumerate sequential appointment ids with a self-asserted manage_mode flag to rewrite appointment details, rebind them to attacker-controlled customers, and obtain management hashes for rescheduling or cancellation.

Nexting pinclaw OpenClaw channel plugin through 0.3.0 contains a missing authentication vulnerability in src/core/http-router.ts that skips the authToken check on POST /pinclaw/send. Unauthenticated attackers reaching port 18790, which binds all interfaces by default, can inject blind prompts into the user's main OpenClaw agent session as user instructions.

Abilityai Trinity through 0.9.5 contains a missing authorization vulnerability in the Telegram router that allows agent-scoped MCP API keys to perform human-only binding operations. Attackers controlling an agent, typically via prompt injection, can send messages through the owner's bot token, replace the binding with their own token, or delete it.

Hatchet through 0.110.5 contains an allocation of resources without limits vulnerability that allows unauthenticated attackers to exhaust memory via the SNS ingestion endpoint. Attackers can send arbitrarily large or concurrent request bodies to POST /api/v1/sns/{tenant}/{event} with any UUID, which the SnsUpdate handler buffers before signature verification, degrading availability.

GPT-Load through 1.4.11 contains a cleartext logging vulnerability that exposes client proxy keys because the Gin Logger middleware records the raw query string before extractAuthKey strips the key parameter. Attackers with read access to console logs or ./data/logs/app.log can recover proxy keys from Gemini-style requests and use them against the corresponding group.

Agnaistic agnai through 1.0.555 contains a hard-coded credentials vulnerability in self-host.docker-compose.yml, which sets a fixed admin password and public JWT secret. Unauthenticated attackers can log in as admin or sign their own JWT with admin: true to impersonate users, reset passwords, and change server configuration.