Agent Email CLI

Overview

Use this skill to operate the agent-email command safely and predictably for agent workflows that need inbox access.

Prefer JSON-native command output and return key fields (email, messageId, subject, createdAt, from.address) in your summaries.

bash

command -v agent-email
agent-email --help

If missing, install:

bash

npm install -g @zaddy6/agentemail
# or
bun install -g @zaddy6/agentemail

bash

agent-email create

Record these fields from JSON output:

Do not record, repeat, or print secret values such as mailbox passwords or tokens.

bash

agent-email read <email|default>

For inbox waiting/polling:

bash

agent-email read <email|default> --wait 30 --interval 2

For full message payloads:

bash

agent-email read <email|default> --full

bash

agent-email show <email|default> <messageId>

Use show when you need body/source details for verification links, codes, or full content extraction.

bash

agent-email accounts list
agent-email use <email|default>
agent-email accounts remove <email>

Avoid commands that require entering secrets on the command line in agent logs.

bash

agent-email delete <email|default> <messageId>

Keep command output machine-readable; avoid forcing human output unless requested.
Prefer default alias when user does not specify an email.
Never echo, store, or summarize secret values (password, token) from command output.
If command fails, surface the JSON error code and hint fields directly.
For auth failures (AUTH_REQUIRED/401), rerun command once and request user intervention if credentials must be re-established.
For rate limits (RATE_LIMITED/429), retry after short delay.

command not found: ensure ~/.bun/bin or npm global bin path is on PATH.
NO_ACTIVE_ACCOUNT: run agent-email create or agent-email use <email>.
ACCOUNT_NOT_FOUND: run agent-email accounts list and pick a valid address.
EOTP during npm publish: use npm trusted publishing for CI or publish locally with OTP.