Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed AgentMailer email routing skill with expected email read/write capabilities and no hidden local code or persistence.
Install only if you intend to let AgentMailer handle mailbox data and user-directed email actions, including drafts, sends, scheduling, labels, and deletions. Use focused read-only or send workflows where possible and require clear confirmation before external sends or destructive mailbox changes.
The manifest presents this skill as a routing-only component, but it exposes a backend tool capable of reading, drafting, and sending email. That mismatch violates least privilege and can let a supposedly non-mutating router invoke higher-risk operations if routing logic is bypassed, confused, or abused by downstream prompts.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
value: "agentmailer"
description: "Read, draft, and send AgentMailer email"
transport: "streamable_http"
url: "https://api.agentmailer.ai/mcp"
policy:
allow_implicit_invocation: false
Drafting is broader than the stated purpose of merely routing requests and increases the available action surface unnecessarily. Even if drafting is not as severe as sending, it can expose message content handling, persistence, or preparatory actions that a routing-only skill should not perform.
No suspicious patterns detected.