Vulnerabilities
All
My stack
Known exploited
Public exploit
Critical (9+)
Likely exploited (EPSS ≥ 50%)
Changed, 24 h
New, 24 h
Sort:
risk
modified
published
CVSS
EPSS
| CVE | CVSS | EPSS | Vendor, product | Description | Modified |
|---|---|---|---|---|---|
| CVE-2026-110850 | 7.2 high | 0.00% | no CPE yet | n8n before 2.41.4 and 2.42.0 before 2.42.1 does not validate credential ids stored in agent node parameters when an editor saves a shared workflow, because the check inspects only a node's credentials field. An editor can configure a Message an Agent node in inline mode with tools they control to resolve the owner's credentials and send decrypted secrets to an attacker-chosen destination. | 15 Oct |
| CVE-2026-96667 | 7.2 high | 0.00% | no CPE yet | The Real Estate Manager – Property Listing and Agent Management plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'first_name' parameter in all versions up to, and including, 7.3 due to insufficient input sanitization and output escaping. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. The reCAPTCHA check is trivially bypassed by omitting the g-recaptcha-response parameter entirely, since validation only runs when that parameter is present. | 13 Oct |
| CVE-2025-11204 | 7.2 high | 0.00% | no CPE yet | The RegistrationMagic – Custom Registration Forms, User Registration, Payment, and User Login plugin for WordPress is vulnerable to SQL Injection in all versions up to, and including, 6.0.6.2 due to insufficient escaping on the user supplied parameter and lack of sufficient preparation on the existing SQL query. This makes it possible for authenticated attackers, with administrator access or higher, to append additional SQL queries into already existing queries that can be used to extract sensitive information from the database. An unauthenticated attacker could utilize an injected Cross-Site Scripting via user-agent on form submission to leverage this to achieve Reflected Cross-Site Scripting. | 8 Oct |
| CVE-2026-119337 | 7.1 high | — | infiniflow, ragflow | RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine. From 0.25.2 until 0.27.3, POST /api/v1/agents/chat/completions invokes completion_openai and returns through the openai-compatible branch before UserCanvasService.accessible() checks whether the caller may access the requested agent. An authenticated user who knows another user's private agent UUID can execute that Only me workflow and read its output across the user or tenant boundary; configured tools may also cause side effects, but stronger impacts were not demonstrated. This issue is fixed in version 0.27.3. | 21 h ago |
| CVE-2026-116242 | 7.1 high | — | pgadmin.org, pgadmin 4 | pgAdmin 4 opens SSH tunnels with the sshtunnel library but gave it neither host_pkey_directories nor ssh_config_file, so it always loaded the unencrypted private keys in ~/.ssh of the operating-system account running pgAdmin, read that account's ~/.ssh/config (IdentityFile, User, HostName, ProxyCommand) for the tunnel host, and could use its ssh-agent. In server mode any user could therefore configure a tunnel to a host that accepts the service account's key, with no credentials of their own or with an identity file that did not exist in their storage, and reach it as that account. The fix makes tunnels in server mode use only the identity file or password the user configured: the service account's keys, configuration and agent are no longer used, and an identity file that cannot be found in the user's storage is reported instead of falling back to ~/.ssh. Desktop mode is unchanged. This issue affects pgAdmin 4: before 9.19. | 21 h ago |
| CVE-2025-30414 | 7.1 high | 0.00% | no CPE yet | Local privilege escalation due to improper authentication. The following products are affected: Acronis Cyber Protect Cloud Agent (Windows) before build 42569, Acronis Cyber Protect 17 (Windows) before build 42587. | Wed 28 Oct |
| CVE-2026-118038 | 7.1 high | 0.00% | no CPE yet | In Progress® Telerik® Document Processing Libraries prior to version 2026.3.1027, a path traversal vulnerability in the Convert and Merge AI agent tools allows an authenticated attacker with low privileges who can supply tool arguments to read files outside the configured output directory, and to overwrite such files where they are writable, via relative file paths that resolve against the process current working directory. | Wed 28 Oct |
| CVE-2026-39308 | 7.1 high | 0.00% | praison, praisonai | PraisonAI is a multi-agent teams system. Prior to 1.5.113, PraisonAI's recipe registry publish endpoint writes uploaded recipe bundles to a filesystem path derived from the bundle's internal manifest.json before it verifies that the manifest name and version match the HTTP route. A malicious publisher can place ../ traversal sequences in the bundle manifest and cause the registry server to create files outside the configured registry root even though the request is ultimately rejected with HTTP 400. This is an arbitrary file write / path traversal issue on the registry host. It affects deployments that expose the recipe registry publish flow. If the registry is intentionally run without a token, any network client that can reach the service can trigger it. If a token is configured, any user with publish access can still exploit it. This vulnerability is fixed in 1.5.113. | Mon 26 Oct |
| CVE-2026-40160 | 7.1 high | 0.00% | praison, praisonaiagents | PraisonAIAgents is a multi-agent teams system. Prior to 1.5.128, web_crawl's httpx fallback path passes user-supplied URLs directly to httpx.AsyncClient.get() with follow_redirects=True and no host validation. An LLM agent tricked into crawling an internal URL can reach cloud metadata endpoints (169.254.169.254), internal services, and localhost. The response content is returned to the agent and may appear in output visible to the attacker. This fallback is the default crawl path on a fresh PraisonAI installation (no Tavily key, no Crawl4AI installed). This vulnerability is fixed in 1.5.128. | Mon 26 Oct |
| CVE-2026-40518 | 7.1 high | 0.00% | bytedance, deerflow | ByteDance DeerFlow before commit 2176b2b contains a path traversal and arbitrary file write vulnerability in bootstrap-mode custom-agent creation where the agent name validation is bypassed. Attackers can supply traversal-style values or absolute paths as the agent name to influence directory creation and write files outside the intended custom-agent directory, potentially achieving arbitrary file write on the system subject to filesystem permissions. | Mon 26 Oct |
| CVE-2026-42261 | 7.1 high | 0.00% | legeling, prompthub | PromptHub is an all-in-one AI toolbox for prompt, skill, and agent management. From version 0.4.9 to before version 0.5.4, apps/web/src/routes/skills.ts exposes an authenticated endpoint POST /api/skills/fetch-remote that fetches a user-supplied URL server-side and reflects the response body (up to 5 MB) back to the caller. The SSRF protection in apps/web/src/utils/remote-http.ts (isPrivateIPv6) attempts to block private/loopback destinations, but multiple alternate-but-valid IPv6 representations bypass the check. The bypasses reach any IPv4 address (loopback, RFC1918, link-local) via IPv4-mapped IPv6 in hex form, and the canonical ::1 via any representation that isn't the literal string "::1". Any authenticated user (role: user or admin) can trigger the SSRF. On deployments configured with ALLOW_REGISTRATION=true — a supported and documented configuration — this means any internet user who can register. This issue has been patched in version 0.5.4. | Mon 26 Oct |
| CVE-2026-41287 | 7.1 high | 0.00% | watchguard, agent | Stack-based Buffer Overflow vulnerability in the WatchGuard Agent discovery service on Windows allows Overflow Buffers. An unauthenticated attacker on the same local network could exploit this vulnerability to crash the agent service. | Mon 26 Oct |
| CVE-2026-41286 | 7.1 high | 0.00% | watchguard, agent | Stack-based Buffer Overflow vulnerability in the WatchGuard Agent discovery service on Windows allows Overflow Buffers. An unauthenticated attacker on the same local network could exploit this vulnerability to crash the agent service. | Mon 26 Oct |
| CVE-2026-74047 | 7.1 high | 0.00% | no CPE yet | Wazuh from 4.0.0 before 4.14.8 contains an information disclosure vulnerability that allows low-privileged users with agent:read to retrieve the plaintext authd.pass enrollment password via the unmasked agent configuration endpoint. Attackers holding the default readonly role can request the auth component configuration for pseudo-agent ID 000 to obtain the secret and enroll rogue agents against the manager. | Mon 26 Oct |
| CVE-2026-74045 | 7.1 high | 0.00% | no CPE yet | Wazuh from 3.12.0 before 4.14.8 contains a heap buffer overflow in fim_generate_alert() where full_log is sized from input but written bounded by OS_MAXSTR. An enrolled agent can send a crafted FIM syscheck event to corrupt manager heap memory and deterministically crash wazuh-analysisd, causing a fleet-wide detection outage. | Mon 26 Oct |
| CVE-2026-47118 | 7.1 high | 0.00% | no CPE yet | Agent Zero before version 1.15 contains a path traversal vulnerability that allows unauthenticated attackers to read arbitrary files by supplying crafted paths to the image file serving endpoint, which relies solely on an extension allowlist while the path containment check is explicitly disabled. Attackers can request any file with an image extension readable by the process, including files outside the agent workspace, user home directories, and mounted volumes, and can also leverage symlink-based escapes due to the lack of path canonicalization in the path resolution logic. | Mon 26 Oct |
| CVE-2026-50141 | 7.1 high | 0.00% | no CPE yet | Woodpecker is a CI/CD engine. Starting in version 3.0.0 and prior to version 3.14.1, a vulnerability in Woodpecker CI's gRPC layer allowed any authenticated agent to impersonate any other agent on the same server by injecting a forged `agent_id` value into outgoing gRPC metadata. The server correctly verified the JWT token but then discarded the verified agent identity in favor of the client-supplied value. Version 3.14.1 patches the issue. As a workaround, disable org agents (`WOODPECKER_DISABLE_USER_AGENT_REGISTRATION=true`) and delete existing ones. | Sun 25 Oct |
| CVE-2026-63175 | 7.1 high | 0.00% | no CPE yet | PlaywrightCapture stored capture-specific configuration and runtime data as mutable class-level variables rather than instance-level variables. Consequently, multiple Capture objects running within the same Python process could share state, including HTTP headers, cookies, browser storage, HTTP credentials, proxy configuration, user-agent settings, geolocation information, and captured request data. In a multi-user or concurrent deployment, information supplied during one capture could therefore persist and be reused by a subsequent or parallel capture. This could result in the disclosure of authentication cookies, credentials, browser storage, or captured request data belonging to another user. It could also cause requests to be performed with another capture's authentication context, headers, or proxy configuration, potentially enabling unauthorized access to remote resources or interference with other capture operations. The vulnerability is resolved by initializing all capture-specific settings and request data as instance variables in the Capture constructor, ensuring that state is isolated between capture operations. | Sat 24 Oct |
| CVE-2026-58484 | 7.1 high | 0.00% | network-ai | Network-AI is a TypeScript/Node.js multi-agent orchestrator. Prior to version 5.12.2, `EnvironmentManager.listBackups()` reads each backup's `_manifest.json` and trusts the manifest's `path` field. `EnvironmentManager.pruneBackups()` later passes that trusted `entry.path` directly to `rmSync(entry.path, { recursive: true, force: true })`. An attacker who can place or modify a manifest inside `data/<env>/.backups/<name>/_manifest.json` can cause `network-ai env backup prune --env <env> --keep <n>` or any code path invoking `pruneBackups()` to recursively delete an arbitrary path accessible to the Network-AI process user. This is fixed in v5.12.2. `pruneBackups()` no longer passes `entry.path` from the on-disk manifest to `rmSync`. The deletion path is recomputed from a format-validated `entry.backupId`, and a `dirname` containment check confines deletion to exactly one level under the backups directory. A poisoned manifest (e.g. `"path": "/"`) is now inert. | Sat 24 Oct |
| CVE-2026-57494 | 7.1 high | 0.00% | no CPE yet | AgenticMail gives AI agents real email addresses and phone numbers. In @agenticmail/api prior to version 0.9.64, a low-privileged authenticated AgenticMail agent can enumerate another agent's pending/claimed tasks by supplying the target agent name to `GET /api/agenticmail/tasks/pending?assignee=<name>`. The returned task objects include the task IDs and payloads. The same task IDs can then be used with the capability-style task mutation endpoints (`/tasks/:id/claim`, `/tasks/:id/result`, `/tasks/:id/complete`, `/tasks/:id/fail`) to claim, complete, or fail tasks assigned to a different agent. Because ordinary authenticated agents can discover agent names through `GET /api/agenticmail/accounts/directory`, the task ID effectively stops being a secret capability. This turns the intended capability model into a cross-agent authorization bypass. Version 0.9.64 contains a fix. | Sat 24 Oct |
| CVE-2026-47397 | 7.1 high | 0.00% | no CPE yet | PraisonAI is a multi-agent teams system. Prior to version 4.6.40, hidden metadata in a webpage causes PraisonAI agents to write attacker-controlled content to arbitrary paths. `write_file` skips path validation when `workspace=None` (always `None` in production). Version 4.6.40 fixes the issue. | Sat 24 Oct |
| CVE-2026-54715 | 7.1 high | 0.00% | no CPE yet | GoAccess is a real-time web log analyzer and interactive viewer that runs in a terminal in *nix systems or through the browser. In version 1.10.2, parse_browser assumes the matched browser token begins with Opera and moves a trailing version substring to match plus five, allowing a crafted User-Agent in a processed access log to write one to four attacker-influenced bytes beyond the heap allocation and corrupt or crash GoAccess. This issue is fixed in version 1.11. | 23 Oct |
| CVE-2026-111384 | 7.1 high | 0.00% | no CPE yet | Nezha 1.0.0 before 2.3.15 contains a memory exhaustion vulnerability in the dashboard file manager WebSocket endpoint because fmStream never calls SetReadLimit, leaving message size unbounded. Authenticated users with file manager scopes can send one oversized binary message to /api/v1/ws/file/:id, triggering an OOM kill that drops all agent monitoring. | 23 Oct |
| CVE-2026-93336 | 7.1 high | 0.00% | no CPE yet | Missing authorization in the unit resources download route in Canonical Juju 4.0 through 4.0.15 allows a controller-authenticated principal (a unit agent, or a user with only login permission) to read any application's charm resource in any model, including container-registry credentials, via a crafted request naming an arbitrary unit. | 23 Oct |
| CVE-2026-40145 | 7.1 high | 0.00% | no CPE yet | A vulnerability exists in the interaction between a Endpoint Privilege Management (Windows Deployment) support utility and the agent's tamper protection controls. Under certain conditions, the protections applied to the utility process may not be enforced as intended. | 23 Oct |
| CVE-2026-115478 | 7.1 high | 0.00% | no CPE yet | obsidian-tc is a governed, agent-ready Model Context Protocol server for Obsidian vaults. From 1.0.1 until 1.31.4, `create_entity`, `link_entities`, `rename_entity`, `unlink_entities` and `delete_entity` regenerate memory notes from SQLite at the computed `memory/<type>/<name>.md` path without reliably preserving note ownership or the consistency of a failed lifecycle operation. During shared materialization, a hand-authored note or an orphan left by a different entity whose name sanitized to the same path could have its entire body overwritten on the next write naming that type and name, with non-owned frontmatter preserved. `rename_entity` also seeded the destination path with a direct atomic write without validating ownership and could overwrite a foreign note there. Separately, several tools mutated SQLite before fallible filesystem work: `create_entity` and `link_entities` inserted the row or relation before materialization; `rename_entity` updated the entity, including a retired status, before updating its note; `unlink_entities` removed a relation before rematerializing the source note; and `delete_entity`'s cascade removed the target and its relations before rematerializing neighbouring notes. A later filesystem failure could therefore leave a row or relation changed and the memory graph inconsistent with the notes on disk even when the operation failed. An authorized caller with the required memory-tool permissions could, without intending to, destroy the body of a vault note outside the entity being written; `delete_entity` additionally requires its deletion permission and destructive confirmation, which this memory flaw does not bypass. Snapshots-before-write through `restore_note` mitigate the overwrite only where snapshots are enabled and the note had previously been written through obsidian-tc. This issue is fixed in version 1.31.4. | 23 Oct |
| CVE-2026-115477 | 7.1 high | 0.00% | no CPE yet | obsidian-tc is a governed, agent-ready Model Context Protocol server for Obsidian vaults. From 1.13.0 until 1.31.4, on a Streamable HTTP deployment with `auth.jwtSecret` configured and a 2026-07-28-era client declaring `elicitation.form` support, a dispatch-gated tool call offered an `inputRequired` confirmation could still complete after the human declined or cancelled, or the elicitation response was omitted or malformed, provided the client echoed the validly bound `requestState` on resubmission. The server trusted an authentic, call-bound state as proof of approval even though that state was minted before the `elicitation/create` leg was asked and proved only that the resubmission belonged to the same call. A client behaving incorrectly or under an attacker's control could therefore execute dispatch-gated destructive vault operations without the human's approval, while still requiring the caller's ordinary tool permissions. Only tools marked `destructive: true` or requiring a scope with a human-in-the-loop floor were reachable through this bypass. The sixteen handler-side conditionally gated tools, including `write_note` overwrite and `move_note`, did not read elicitation state on the modern wire and let neither approval nor decline through that path. Affected deployments using stdio or leaving `auth.jwtSecret` unset instead used the CLI `elicit_token` fallback and did not offer this `requestState` round trip. The fix does not prevent a malicious client from reporting a fabricated approval, because the server cannot verify that the client displayed the form or that a human saw it. This issue is fixed in version 1.31.4. | 23 Oct |
| CVE-2026-114067 | 7.1 high | 0.00% | no CPE yet | Trigger.dev 4.5.0 before 4.6.2 contains an improper authorization vulnerability in the PUT snapshot-url session endpoint that allows public session token holders to obtain presigned S3 write URLs. End users can overwrite their own session's durable chat snapshot to forge the agent's conversation history, since the route lacks a secret-key-only check. | 22 Oct |
| CVE-2026-68425 | 7.1 high | 0.00% | no CPE yet | In the Linux kernel, the following vulnerability has been resolved: IB/mad: Drop unmatched RMPP responses before reassembly Kernel-handled RMPP receive processing starts reassembly for active DATA responses before the response is matched to an outstanding send. The normal match happens later, after ib_process_rmpp_recv_wc() has either assembled a complete message or consumed the segment. That ordering lets an unsolicited response that routes to a kernel RMPP agent by the high TID bits allocate or extend RMPP receive state before the full TID and source address are checked against a real request. A reordered burst can therefore reach the receive-side insertion path even though the response would not match any send. For kernel-handled RMPP DATA responses, require the existing ib_find_send_mad() match before entering RMPP reassembly. The matcher already checks the full TID, management class and source address/GID against the agent wait, backlog and in-flight send lists. If there is no match, drop the response without creating RMPP state. This leaves the RMPP window behavior unchanged and only rejects responses that have no corresponding request. | 19 Oct |
| CVE-2025-64168 | 7.1 high | 0.00% | no CPE yet | Agno is a multi-agent framework, runtime and control plane. From 2.0.0 to before 2.2.2, under high concurrency, when session_state is passed to Agent or Team during run or arun calls, a race condition can occur, causing a session_state to be assigned and persisted to the incorrect session. This may result in user data from one session being exposed to another user. This has been patched in version 2.2.2. | 19 Oct |
| CVE-2025-10918 | 7.1 high | 0.00% | ivanti, endpoint_manager | Insecure default permissions in the agent of Ivanti Endpoint Manager before version 2024 SU4 allows a local authenticated attacker to write arbitrary files anywhere on disk | 18 Oct |
| CVE-2026-55787 | 7.1 high | 0.00% | no CPE yet | Flyto2 Core is an execution kernel for automation and AI-agent workflows. Prior to 2.26.3, validate_url_ssrf and is_private_ip in src/core/utils.py compare resolved addresses only against native private IPv4 and IPv6 ranges and do not unwrap IPv4-mapped, IPv4-compatible, 6to4, or NAT64 transition addresses. An authenticated workflow author can provide a transition-form literal URL through the Execution API POST /v1/execute route, http.get, or sibling guarded outbound-request modules, causing the guard to accept an address whose embedded IPv4 points to loopback, RFC 1918, or cloud metadata services. On a host that routes the selected transition form, the outbound request reaches the internal destination and returns the response body to the workflow caller, enabling read SSRF and disclosure of internal-service data or cloud credentials. This issue is fixed in version 2.26.3. | 16 Oct |
| CVE-2026-108757 | 7.1 high | 0.00% | no CPE yet | 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. | 16 Oct |
| CVE-2026-111383 | 7.1 high | 0.00% | no CPE yet | Nezha from 2.0.13 before 2.3.15 contains a resource exhaustion vulnerability that allows authenticated users to open unbounded concurrent connections to the /api/v1/ws/transfer WebSocket endpoint. Attackers can hold thousands of subscriptions created by ServerTransferClass.Subscribe without limits, exhausting file descriptors, memory, and goroutines and blocking agent gRPC reconnections. | 16 Oct |
| CVE-2026-96615 | 7.1 high | 0.00% | no CPE yet | Description NGINX Gateway Fabric has a vulnerability when it is configured with Gateway Custom Resource Definition (CRD) and have an authenticated gRPC connection with NGINX Agent. A namespaced authenticated attacker with access to a service account may impersonate the authenticated NGINX Agent gRPC connection to read TLS private keys of Gateway deployments in another namespace. Impact A remote authenticated attacker may be able to read TLS private keys of Gateway deployments in another namespace. There is no data plane exposure; this is a control plane issue only. Note: Software versions which have reached End of Technical Support (EoTS) are not evaluated. | 15 Oct |
| CVE-2026-62676 | 7.1 high | 0.00% | no CPE yet | Omnigent is an open-source AI agent framework and meta-harness for orchestrating coding agents. Prior to 0.3.0, the shared shell-command parser in omnigent/policies/builtins/_shell.py fails to recognize combined interpreter flags, the timeout, nice, setsid, and stdbuf wrappers, command substitutions, and a single background control operator. A gated git push or gh write hidden with these forms produces no parsed operation, causing the github.py write_repos and write_branches allowlist and the working_dir.py workspace confinement policies to abstain and allow the command. An authenticated or prompt-injected agent can therefore push to an unauthorized repository or branch or escape the intended workspace. This issue is fixed in version 0.3.0. | 14 Oct |
| CVE-2026-108883 | 7.1 high | 0.00% | no CPE yet | JeecgBoot through 3.9.5 contains a missing authorization vulnerability in the editThirdAppConfig handler that allows any authenticated user to modify third-party application configurations. Low-privileged attackers can replace client id, client secret, agent id and corp id of DingTalk, WeCom or Feishu integrations to redirect directory synchronisation and messaging to attacker-controlled applications or break them. | 13 Oct |
| CVE-2026-106145 | 7.1 high | 0.00% | no CPE yet | In Progress® Telerik® Report Server prior to version 12.2.26.1007, incorrect privilege assignment in the service-agent SignalR hub allows an authenticated user, including a low-privilege or guest account with a valid bearer token, to register as a trusted service agent. On the next server settings-synchronization event, the rogue agent receives storage settings and encryption private keys. This privilege escalation enables disclosure of protected secrets, including stored data-source credentials and connection strings, and allows agent impersonation and interference with task dispatch. | 9 Oct |
| CVE-2026-100582 | 7.1 high | 0.00% | no CPE yet | OpenClaw channel plugins (@openclaw/msteams, @openclaw/feishu, @openclaw/matrix, and @openclaw/googlechat) before 2026.8.1 do not enforce the configured channel read allowlist for caller-supplied explicit read targets in message, reaction, pin, member, and related metadata read actions. A lower-trust sender or a steered agent with access to a channel read action can therefore retrieve content or metadata from channels or rooms excluded by the operator's read policy; the practical impact depends on the permissions held by the connected bot account. The issue is fixed in 2026.8.1. | 8 Oct |
| CVE-2026-77897 | 7.0 high | 0.00% | microsoft, power automate agent for virtual desktops, power automate for desktop | Relative path traversal in Power Automate allows an authorized attacker to elevate privileges locally. | 19 h ago |
| CVE-2026-54321 | 7.0 high | 0.00% | no CPE yet | Daytona is a secure and elastic infrastructure runtime for AI-generated code execution and agent workflows. From 0.101.0 until 0.184.0, sandbox previews that were switched from public to private could remain reachable without authentication for a short period after the change, due to a cached visibility state that was not invalidated when the sandbox's visibility changed. This vulnerability is fixed in 0.184.0. | Sun 25 Oct |
| CVE-2026-44934 | 7.0 high | 0.00% | no CPE yet | A information disclosure when DEBUG loglevel is set in SUSE Rancher AI Agent 1.0 before 1.0.2 could leak API keys or LLM response text with potential sensitive data into logfiles, allowing local attackers to misuse respective gained data or credentials. | Sun 25 Oct |
| CVE-2026-72718 | 7.0 high | 0.00% | no CPE yet | goose is general-purpose AI agent that runs on your machine. Prior to 1.44.0, the `goose review` command runs the system `git` executable to gather the diff for review without stripping attacker-controlled Git configuration. A malicious repository whose `.git/config` sets [`core] fsmonitor = <command>` causes Git to execute that command on the host during the index refresh performed by `git diff HEAD`. The command runs before goose contacts a model and without a submitted prompt, model call, tool approval, or trust prompt. The context-gathering Git process is not sandboxed and is outside goose's tool-permission model. Arbitrary commands run with the privileges and environment of the user running goose, allowing file access or modification and exfiltration of environment secrets and provider API keys. The vulnerable Git invocations are built by git_command() in crates/goose-cli/src/commands/review/handler.rs and are used by touched_files() and collect_diff() for `git diff --name-only HEAD` and `git diff HEAD`. This issue is fixed in version 1.44.0. | 24 Oct |
| CVE-2025-61667 | 7.0 high | 0.00% | no CPE yet | The Datadog Agent collects events and metrics from hosts and sends them to Datadog. A vulnerability within the Datadog Linux Host Agent versions 7.65.0 through 7.70.2 exists due to insufficient permissions being set on the `opt/datadog-agent/python-scripts/__pycache__` directory during installation. Code in this directory is only run by the Agent during Agent install/upgrades. This could allow an attacker with local access to modify files in this directory, which would then subsequently be run when the Agent is upgraded, resulting in local privilege escalation. This issue requires local access to the host and a valid low privilege account to be vulnerable. Note that this vulnerability only impacts the Linux Host Agent. Other variations of the Agent including the container, kubernetes, windows host and other agents are not impacted. Version 7.71.0 contains a patch for the issue. | 18 Oct |
| CVE-2026-119053 | 6.9 medium | 0.00% | lwip | lwIP 2.0.0 through 2.2.1 built with the SNMP agent contains an out-of-bounds read vulnerability in snmp_pbuf_stream_writeto() that bounds copies by pbuf->len instead of remaining bytes. Remote unauthenticated attackers knowing the read community can send fragmented SNMP requests to receive adjacent receive-buffer or pool memory echoed in responses. | Thu 29 Oct |
| CVE-2026-5131 | 6.9 medium | 0.00% | no CPE yet | GREENmod uses named pipes for communication between plugins, the web portal, and the system service, but the access control lists for these pipes are configured incorrectly. This allows an attacker to communicate with the stream and upload any XML or JSON file, which will be processed by the named pipe with the privileges of the user under whose context the service is running. This allows for Server-Side Request Forgery to any Windows system on which the agent is installed and which provides communication via SMB or WebDav. This issue was fixed in version 2.8.33. | Mon 26 Oct |
| CVE-2026-61428 | 6.9 medium | 0.00% | no CPE yet | PraisonAI AgentMail versions before 4.6.78 lack signature verification in webhook mode, allowing unauthenticated attackers to inject messages with spoofed sender addresses. Attackers can POST crafted message.received events to the webhook endpoint to inject arbitrary content into the agent and trigger replies to attacker-controlled addresses, bypassing sender allow/block lists. | Sun 25 Oct |
| CVE-2026-60089 | 6.9 medium | 0.00% | no CPE yet | PraisonAI (pip package praisonaiagents) before 1.6.78 automatically loads defaults from a project-local .praisonai/config.toml when constructing an Agent, and does not validate the defaults.output.output_file path. A repository-controlled config file can set output_file to an absolute or '..' traversal path; when the developer subsequently calls agent.start() without explicitly passing an output parameter, PraisonAI writes the agent response to that path (creating parent directories as needed), allowing an untrusted checked-out project to overwrite files outside the project root with the privileges of the user running PraisonAI. | Sat 24 Oct |
| CVE-2026-15746 | 6.9 medium | 0.00% | no CPE yet | Strands Agents is an open-source Python SDK for building and running AI agents. The strands-agents-tools package provides pre-built tools for use with the SDK, including the elasticsearch_memory tool for agent memory storage. We identified CVE-2026-15746, a server-side request forgery (SSRF) issue in the elasticsearch_memory tool. The tool exposed its connection parameters (es_url, cloud_id, api_key) as fields the large language model (LLM) could control through the tool schema. When a caller omitted the api_key parameter, the tool fell back to the operator's ELASTICSEARCH_API_KEY environment variable and sent it to whichever host the LLM specified. A crafted prompt could cause the tool to connect to a threat-actor-controlled server and disclose the operator's Elasticsearch API key in the Authorization header. We recommend you upgrade to strands-agents-tools version 0.7.0 or later. As a precautionary measure, we recommend all operators rotate their ELASTICSEARCH_API_KEY, even if there is no indication the credential was exposed. | Sat 24 Oct |
| CVE-2026-115378 | 6.9 medium | 0.00% | no CPE yet | Incus before 6.0.7, 6.1.0 before 7.0.2, and 7.1.0 before 7.5.0 contains an arbitrary file write vulnerability in sftpRecursivePullFile that allows a malicious incus-agent to escape the target directory during incus file pull -r. Attackers with root in a virtual machine can return duplicate entries reported first as symlinks then as files or directories, writing attacker-controlled files such as /etc/shadow on the operator's host. | 23 Oct |
Download these 806 as CSV · subscribe to this list.
EPSS scores: 2026-10-30. NVD cursor: 2026-10-29 00:22. The News page shows only notable CVEs (KEV, your stack, new criticals, EPSS over 50%, or active exploitation per SSVC); this table has all of them.