6 vulnerabilities across 3 products scored HIGH or above on October 23, 2026.

  • 🟠 HIGH: 6

🟠 [HIGH] mervinpraison/praisonai

4 CVEs | CVSS 3.1: 9.8 | AAS 10.8

  • cpe:2.3:a:mervinpraison:praisonai:*:*:*:*:*:*:*:*

PraisonAI, an open-source multi-agent teams framework by mervinpraison, is affected by four vulnerabilities, including at least one critical-severity issue carrying a CVSS 3.1 score of 9.8. The most severe flaw involves unauthenticated access to the AgentOS FastAPI deployment surface, allowing any remote attacker to invoke deployed agents and their associated tools without valid credentials, effectively bypassing prior authentication fixes. Versions 4.2.1 through 4.6.58 are affected.

Organizations running PraisonAI AgentOS deployments, particularly those exposing agent endpoints with tool access, should treat this as high priority. Upgrade to version 4.6.59 or later immediately, and review the vendor advisory at the linked GitHub Security Advisory for details on all four CVEs. Until patched, restrict network access to AgentOS endpoints and audit logs for unauthorized agent invocations.

Vendor Advisory

🟠 [HIGH] openwrt/odhcpd

1 CVE | CVSS 3.1: 9.8 | AAS 10.6

  • cpe:2.3:a:openwrt:odhcpd:*:*:*:*:*:*:*:* (< 25.12)

odhcpd, the DHCPv6/DHCPv4 and router-advertisement daemon used across OpenWrt deployments, contains a critical stack buffer overflow vulnerability with a CVSS 3.1 score of 9.8. A remote unauthenticated attacker can send a single crafted DHCPv6 REQUEST packet to overflow a fixed 512-byte stack buffer via two independent overflow sites, potentially achieving code execution as root, particularly on embedded targets that lack stack canaries. A proof-of-concept exploit is available, increasing the urgency.

All OpenWrt installations running odhcpd versions prior to 25.12 are affected. Network administrators and IoT device operators should upgrade odhcpd to version 25.12 or later immediately, and review the vendor advisory at the linked GitHub Security Advisory for full technical details. Where immediate patching is not feasible, restrict DHCPv6 traffic to trusted network segments.

Vendor Advisory

🟠 [HIGH] fosowl/agenticseek

1 CVE | CVSS 4.0: 9.2 | AAS 10.5

  • cpe:2.3:a:fosowl:agenticseek:*:*:*:*:*:*:*:* (< 409840c)

AgenticSeek, an open-source AI agent framework by Fosowl, contains a remote code execution vulnerability with a CVSS 4.0 score of 9.2. The llm_server component hardcodes Flask debug mode while binding to all network interfaces, allowing any attacker who can reach the server port to derive the Werkzeug debugger PIN and execute arbitrary Python and operating system commands with the server’s privileges.

Deployments running AgenticSeek prior to commit 409840c are affected. Teams using AgenticSeek should update to a commit at or after 409840c immediately, and audit whether the llm_server port has been exposed to untrusted networks. If patching is not immediately possible, restrict network access to the llm_server port and disable debug mode in the Flask configuration.

Vendor Advisory