What problem does it solve? Managing a Feishu (Lark) mailbox through an AI agent requires knowing dozens of API endpoints, permission scopes, and safe sending practices. This Skill provides a complete operational guide for drafting, sending, replying, forwarding, reading, and searching emails through lark-cli, with built-in safeguards against accidental sends and prompt injection from untrusted email content. ## Core Features & Use Cases - Full email lifecycle: Compose new emails, reply, reply-all, and forward with automatic subject prefixes, RFC 2822 threading headers, HTML bodies, attachments, and inline CID images. - Draft-first safety model: All send-type commands save as drafts by default and require explicit user confirmation before actual delivery, plus post-send delivery status verification. - Mailbox management: Triage inbox summaries, full-text search, read messages and threads, manage folders, labels, contacts, download attachments, and watch for incoming mail events via WebSocket. - Use Case: Ask the agent to "reply to the latest email from my boss confirming the schedule" — it reads the message, creates a reply draft with proper threading headers, shows you a summary, and sends only after your approval. ## Quick Start Ask the agent to check your Feishu inbox and draft a reply to the most recent unread email.