CVE-2026-57495
HighCVSS 8.2Exploitation Probability (EPSS)
Low risk20th percentile - higher than 20% of all known CVEs
Summary
Vulnerability in AgenticMail packages (@agenticmail/claudecode before 0.2.39, @agenticmail/codex before 0.1.33, @agenticmail/core before 0.9.43, @agenticmail/openclaw before 0.5.71). An external email sent to the bridge inbox causes the operator's Claude Code session to resume with bypassed permissions, enabling indirect prompt injection into a fully-privileged agent.
Risk Assessment
Risk of an attacker taking control of an AI agent session, potentially leading to unauthorized actions on behalf of the operator.
Recommendation
Upgrade all AgenticMail packages to the fixed versions: @agenticmail/claudecode 0.2.39, @agenticmail/codex 0.1.33, @agenticmail/core 0.9.43, @agenticmail/openclaw 0.5.71.
Other vulnerabilities in AgenticMail
See all- CVE-2026-57494High
Vulnerability in AgenticMail (versions before 0.9.64) allows a low-privileged authenticated agent to enumerate and claim tasks assigned to other agents. Missing authorization in task endpoints.
- CVE-2026-47255High
Multiple weaknesses in AgenticMail before versions 0.9.32 (API) and 0.9.10 (core), related to validation, binding, inactive-agent hour filtering, SQL identifier validation, metadata ownership checks, blocking direct metadata access via raw SQL, outbound worker secret handling, SMTP control character validation, and TLS certificate verification.
Original NVD description (English source)
AgenticMail gives AI agents real email addresses and phone numbers. In @agenticmail/claudecode prior to version 0.2.39, @agenticmail/codex prior to version 0.1.33, @agenticmail/core prior to version 0.9.43, and @agenticmail/openclaw prior to version 0.5.71, two inbound-mail handlers act on a privileged effect without verifying that the sender is the operator, while a sibling handler in the same repo does. The higher-impact one: any external email routed to the bridge inbox causes the dispatcher to resume the operator's Claude Code session with `permissionMode: 'bypassPermissions'`, embedding the attacker-controlled `from`/`subject`/`preview` verbatim into the prompt the resumed agent reads — an indirect prompt injection into a fully-privileged agent (Bash/Write/Edit/WebFetch + the agenticmail MCP toolbelt) running as the operator's OAuth identity. The sibling operator-query email-reply hook gates the same untrusted-From provenance with `isOperatorReplySender(replyFrom, config.operatorEmail)` (fail-closed); the bridge-wake path — a strictly higher-privilege effect — has no equivalent. @agenticmail/claudecode 0.2.39, @agenticmail/codex 0.1.33, @agenticmail/core 0.9.43, and @agenticmail/openclaw 0.5.71 contain a fix.