EXECUTIVE SUMMARY:
Multiple vulnerabilities primarily affect PraisonAI / praisonaiagents, a multi-agent AI framework. The identified security issues include server-side request forgery (SSRF), arbitrary code execution, unsafe workflow handling, and security control bypasses. These flaws originate from improper validation of URLs, redirect handling, and execution of untrusted workflow components, potentially allowing attackers to access internal services, retrieve sensitive resources, or execute unauthorized code within the affected AI agent environment.[/subscribe_to_unlock_form]
EXECUTIVE SUMMARY:
Multiple vulnerabilities primarily affect PraisonAI / praisonaiagents, a multi-agent AI framework. The identified security issues include server-side request forgery (SSRF), arbitrary code execution, unsafe workflow handling, and security control bypasses. These flaws originate from improper validation of URLs, redirect handling, and execution of untrusted workflow components, potentially allowing attackers to access internal services, retrieve sensitive resources, or execute unauthorized code within the affected AI agent environment.[emaillocker id="1283"]
CVE-2026-55526 (CVSS 7.5 — High): A server-side request forgery (SSRF) vulnerability in PraisonAI's web_crawl tool allows attackers to bypass URL validation through redirects and access internal resources, cloud metadata endpoints, and private services.
CVE-2026-55522 (CVSS 7.8 — High): A code execution vulnerability in PraisonAI workflow include functionality allows attackers to execute arbitrary Python code by importing and running an untrusted tools.py file from an included recipe without enforcing security controls.
CVE-2026-55523 (CVSS 7.7 — High): A vulnerability allows attackers to bypass SSRF protections in web_crawl by using crafted URLs that redirect requests to internal resources. The validation mechanism checks only the supplied URL and not the final redirected destination.
CVE-2026-55528 (CVSS 8.2 — High): An authentication bypass vulnerability in PraisonAI's AgentServer allows unauthenticated attackers to access server endpoints, inject events, and subscribe to event streams because the configured auth_token is not enforced.
CVE-2026-55526 (CVSS 8.5 — High): A server-side request forgery (SSRF) vulnerability in PraisonAI's spider_tools allows attackers to bypass host validation using DNS-based hostnames that resolve to private or loopback IP addresses, enabling access to internal services and cloud metadata resources.
CVE-2026-55524 (CVSS 7.5 — High): A server-side request forgery (SSRF) vulnerability in PraisonAI's web_crawl tool allows attackers to bypass URL validation through HTTP redirects and DNS rebinding, enabling access to internal services and cloud metadata endpoints.
CVE-2026-55530 (CVSS 6.1 — Medium): A missing authorization control vulnerability in PraisonAI's ast_grep_rewrite tool allows attackers to modify files without approval checks, potentially enabling arbitrary code injection through unauthorized file rewrites.
CVE-2026-55527 (CVSS 7.1 — High): A path traversal vulnerability in PraisonAI's FileMemory component allows attackers to use crafted user_id values to write arbitrary JSON files outside the intended memory directory.
RECOMMENDATIONS:
REFERENCES:
The following reports contain further technical details:
https://github.com/advisories/GHSA-5r34-2g38-6569
https://github.com/advisories/GHSA-hxmv-c4g6-5fqc
https://github.com/advisories/GHSA-8hjw-25cg-g52h
https://github.com/advisories/GHSA-7g3p-92qq-8wvh
https://github.com/advisories/GHSA-x44h-65qv-cw74
https://github.com/advisories/GHSA-vg6p-v9vm-6fgj
https://github.com/advisories/GHSA-cfxv-8fw8-rwpv
https://github.com/advisories/GHSA-gxmw-5f7x-6g22