// Security Advisory Triage

PraisonAI accepts unsigned webhooks, allows agent invocation via forged events

GitHub advisories — pip · published 2026-10-08 13:34 UTC
ACTION NEEDED PATCH_NOW SECURITYAPI
WHAT CHANGEDGitHub published CVE-2026-61436 on 7 October 2026. PraisonAI's AgentMail webhook mode accepts unsigned message.received events, allowing an attacker to invoke agents by forging webhook payloads. Patched in version 4.6.78.
WHY IT MATTERSAn unauthenticated attacker can trigger arbitrary agent execution by sending crafted webhook requests to any exposed PraisonAI AgentMail endpoint. This bypasses all intended access controls on agent invocation.
WHAT TO DOCheck if you run PraisonAI with AgentMail webhook mode enabled. If so, upgrade to 4.6.78 immediately. If you cannot upgrade, disable webhook endpoints or place them behind authenticated reverse proxy rules until patched.
Model-reported confidence 95% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+ghsa_id: GHSA-7c92-x8vg-4258
+summary: PraisonAI: AgentMail webhook mode accepts forged unsigned message.received events and invokes agents
+severity: high
+published_at: 2026-10-07T14:07:10Z
+cve_id: CVE-2026-61436
+vulnerabilities[].package.name: praisonai
+vulnerabilities[].first_patched_version: 4.6.78
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.5
triage reasoningAI SDK with authentication bypass in webhook handling allows unauthorized agent invocation.
signalsAI provider SDK affected (PraisonAI agent framework) · authentication bypass: unsigned webhook events accepted · agent invocation via forged messages · high severity rating · patched version available (4.6.78) · new advisory entry
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)