CVE-2026-56837: CWE-287: Improper Authentication in MervinPraison PraisonAI

CVE-2026-56837 describes an improper authentication vulnerability (CWE-287) in MervinPraison's PraisonAI multi-agent system. Specifically, from version 4.6.56 up to but not including 4.6.59, the LinearBot component starts a webhook listener bound to 0.0.0.0 and treats the LINEAR_WEBHOOK_SECRET environment variable as optional. When the secret is absent, the system logs a warning but skips Linear-Signature verification entirely. This allows an unauthenticated attacker who can reach the webhook endpoint to submit forged Linear-Event: AgentSession requests. These forged requests are parsed and scheduled for background processing, eventually invoking BotSessionManager.chat() and causing the bot to post responses back to Linear under the bot token. The proof-of-value demonstrates this can be done offline and deterministically by calling the webhook handler directly and monkey-patching the outbound comment path. The vulnerability is patched in version 4.6.59.

An attacker with network access to the webhook endpoint can bypass authentication and submit forged agent session requests. This leads to unauthorized execution of bot commands and posting of responses under the bot's identity, potentially impacting integrity and availability. Confidentiality impact is low, but integrity is high and availability is low according to the CVSS vector.

Upgrade PraisonAI to version 4.6.59 or later where this vulnerability is patched. Prior to upgrading, ensure that the LINEAR_WEBHOOK_SECRET is configured to enforce signature verification. Without the secret, the webhook listener skips authentication, allowing unauthenticated access. No other mitigations are specified.