stm-assistant

Professional email outreach on behalf of your human. Branded signatures, Reply-To routing, CC tracking.

Safety Notice

This listing is from the official public ClawHub registry. Review SKILL.md and referenced scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "stm-assistant" with this command: npx skills add mouserider/stm-assistant

STM Assistant — Send on Behalf

Send professional emails from a dedicated agent inbox on behalf of your human.

Setup Requirements

  1. Agent inbox — via AgentMail (mcporter call agentmail.create_inbox displayName="Your Assistant") or any SMTP provider
  2. Hosted signature logo — small image (60x60px), hosted URL (Gmail blocks inline base64)
  3. mcporter with agentmail configured

Sending Protocol

Every outbound email MUST:

  1. CC your human — no exceptions. They see every email you send.
  2. Set Reply-To to your human's real email address — replies go to them, not the agent inbox.
  3. Use full legal name on formal correspondence.
  4. Include the branded signature:
<div>
  <div style="font-weight: bold;">Human's Full Name</div>
  <div>human@email.com</div>
  <img src="https://your-hosted-logo.png" width="60" height="60" />
  <div style="color: #aaa; font-size: 10px;">Sent on behalf of [Name] by their personal assistant</div>
</div>

Approval Rules

  • Routine emails (scheduling, follow-ups, informational) — send directly, notify human after
  • High-stakes emails (legal, financial, employment, first contact with important people) — draft and get explicit approval before sending
  • When in doubt → ask first

Sending via AgentMail

mcporter call agentmail.send_message \
  inboxId=your-agent@agentmail.to \
  to='[{"email":"recipient@example.com","name":"Recipient Name"}]' \
  subject="Subject Line" \
  html="<p>Email body with signature</p>" \
  cc='[{"email":"human@email.com"}]'

Important: Use text (plain text) or html (formatted) parameters. Never body or htmlBody — those don't exist and produce empty emails.

Inbox Monitoring

Check for replies on a regular cycle (heartbeat or cron):

mcporter call agentmail.list_threads inboxId=your-agent@agentmail.to limit=5 labels='["unread"]'

After processing a reply:

mcporter call agentmail.update_message \
  inboxId=your-agent@agentmail.to \
  messageId=MSG_ID \
  removeLabels='["unread"]' \
  addLabels='["processed"]'

Always mark processed emails. Unread inbox = broken assistant.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

low-carbon-medicine

低碳生活方式医学咨询。当用户提到低碳饮食、生酮饮食、减肥控糖、糖尿病逆转、代谢综合征、胰岛素抵抗时触发。

Registry SourceRecently Updated
General

x0x-api-smoketest-1777556197822

Scratch skill used to validate CI API publish flow before merge.

Registry SourceRecently Updated
General

java-circular-dependency-breaker

Break circular dependencies in Java multi-module Gradle/Maven projects using interface extraction and business service separation. Triggers: 'circular depend...

Registry SourceRecently Updated
General

Options Trading Brain

Professional options trading intelligence system. Monitors whale flow (Unusual Whales), counts Elliott Waves, analyzes Bollinger Bands, multi-timeframe trend...

Registry SourceRecently Updated