Junglewise Threat Intelligence

CVE-2026-47255: AgenticMail Multiple Vulnerabilities including SQL Injection and Auth Bypass

CVE-2026-47255 · Severity: high · CVSS 8.2 · Published 2026-07-20

Technologies: @agenticmail/api (npm), @agenticmail/core (npm). Vendors: npm.

Executive brief

AgenticMail, a tool that provides AI agents with email and phone capabilities, contained multiple security vulnerabilities. These flaws could have allowed unauthorized access to agent data, the sending of malicious emails, or the bypass of security controls. An attacker could potentially manipulate database queries or intercept communications due to weak encryption defaults and hard-coded credentials.

Technical details

AgenticMail (@agenticmail/api and @agenticmail/core) suffered from a suite of vulnerabilities including SQL injection (CWE-89) via unvalidated 'having', 'where', and 'order-by' clauses in storage routes. It also contained hard-coded secrets for outbound workers (CWE-798), lacked proper SMTP control-character validation (CWE-20), and defaulted to disabled TLS certificate verification (CWE-319). Additionally, improper access controls allowed for potential cross-agent data reads through comma-join SQL queries. These issues are resolved in @agenticmail/api version 0.9.32 and @agenticmail/core version 0.9.10 by implementing strict identifier validation, fail-closed secret handling, and mandatory TLS verification.

Affected products

  • agenticmail @agenticmail/api < 0.9.32
  • agenticmail @agenticmail/core < 0.9.10

Timeline

  • 2026-05-18: patched: Security hardening patch merged into main repository
  • 2026-07-20: disclosed: CVE-2026-47255 published

References

Related threats