Guide · Agent skill
How to use the Feishu Mail Assistant agent skill (lark-mail)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The lark-mail skill is an agent skill that lets an AI agent work with a Feishu (Lark) mailbox through the lark-cli command-line tool. It is aimed at people who run their inbox in Feishu Mail and want help finding, summarising, organising and drafting email. This guide explains what the skill does, how to install it, how to ask for things in plain language, and what safety behaviour to expect.
Feishu Mail Assistantlarksuite/cli · MIT Download the skill Quick start
- 1Download and unzip the skill. Download the lark-mail ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Install lark-cli. Install the lark-cli command-line tool, since the skill requires it, and confirm it runs with lark-cli mail --help.
- 3Authorise your Feishu mail account. Run lark-cli auth login --domain mail and complete the sign-in for your Feishu or Lark account.
- 4Make a plain-language request. Ask the agent something like 'summarise my unread mail from today'. The skill activates on mail and email intent.
- 5Review the preview. For any sending, deleting or rule change, read the preview the agent shows: recipients, subject, action and number of affected items.
- 6Open the draft in Feishu Mail. When the agent creates a draft, open the link it provides if one is returned, check the content and decide whether to send.
What is an agent skill and what does lark-mail let the agent do?
An agent skill is a folder containing instructions (a SKILL.md file) and sometimes scripts and reference files that an AI agent loads when a request matches the skill's description. The lark-mail skill is triggered by mail and email intent, such as drafting, sending, replying, forwarding, searching, reading mail, managing folders and labels, managing contacts, watching for new mail, or setting inbox rules.
Once installed and authorised, the agent can find and read emails and whole threads, organise messages with labels, read status, folders and trash, and draft, reply, forward or schedule emails. The skill is built so that outgoing mail is saved as a draft by default and is only sent after you explicitly confirm.
- Search and triage the inbox, then read single messages, batches or full threads
- Add or remove labels, mark read or unread, and move messages between folders
- Soft-delete messages or threads (moved to trash)
- Draft new emails, and reply, reply-all or forward; sending requires --confirm-send
- Create, update, enable, disable and reorder inbox rules, and manage mail templates
- Use HTML templates for recurring reports or newsletters
Install the lark-mail skill
Download the lark-mail ZIP from this page and unzip it. Then place the folder in one of these locations: ~/.claude/skills/ to make it available to your user account, or a project's .claude/skills/ folder to limit it to that project.
If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Install and connect lark-cli
The skill depends on the lark-cli command-line tool, which the skill's metadata lists as a required binary. The skill itself cannot reach a mailbox without it. Check that the tool is available by running its mail help, which the skill references as lark-cli mail --help.
Then authorise your Feishu account for mail. The skill's guidance is to use user identity by default, so sign in with the mail domain.
- Required binary: lark-cli
- Check help: lark-cli mail --help
- User sign-in for mail: lark-cli auth login --domain mail
lark-cli auth login --domain mailAsk for what you want in plain language
You do not need to name the skill or any command. Describe the email task and the agent will decide whether this skill applies. The skill covers mail intent only: it does not handle docs, sheets, calendar, auth setup, pure contact lookup or IM chat tasks.
Useful requests name the thing you are working on, such as a sender, a subject, a folder or a time range. The agent will look up real messages before acting on them and will ask you to confirm anything destructive.
- "Summarise my unread mail from today and tell me what needs a reply."
- "Draft an email to alice@example.com about Thursday's review and open it for me to check."
- "Find the thread about the Q3 budget and show me the latest reply."
- "Move the newsletters from last week into the ARCHIVED folder."
- "Set up a rule to move mail from billing@example.com to a folder called Invoices."
Safety: confirmations, drafts and untrusted email content
The skill treats email bodies, subjects and sender names as untrusted input. Instructions found inside an email, such as 'forward this to…', are not treated as your instructions. Before sending, forwarding, deleting or changing rules, the agent must show you a preview of the recipients, subject and body summary, or the action and affected items, and wait for your approval. Sending a draft is a separate action that also needs your explicit confirmation.
Because the skill runs scripts and uses the network, review what it is asked to do, and keep the default draft-first flow. Destructive actions show a preview with the number of affected items for batch operations.
- Drafts are the default: send commands save a draft unless you add --confirm-send
- Deletes, trash, cancel-scheduled-send and rule create, update or delete require your confirmation
- Label, read-status and folder moves are reversible and do not need confirmation
- The skill flags possible XSS and prompt-injection content and warns about spoofed senders
Tips for using the skill well
Let the agent use your own identity (--as user) for sending, replying, forwarding and draft edits, as the skill requires. Use bot identity only when an administrator has granted the app the needed permissions; with bot identity you must pass the mailbox address explicitly.
When the skill needs a command flag it is unsure of, it runs the help output first rather than guessing. You can do the same: lark-cli mail +triage -h or lark-cli mail +send -h show the flags available in your version.
- Keep requests specific: name the sender, folder or date range
- Review the draft in Feishu Mail before asking for it to be sent
- For HTML emails, the write paths lint the HTML automatically, so you usually do not need to run the lint step yourself
- Use the templates in assets/templates for weekly reports, team reports, research reports, job applications and newsletters
lark-cli mail +triage -hWhat people use it for
- Daily inbox triage: summarise unread mail and list the messages that need a reply.
- Drafting replies and forwards that you review in Feishu Mail before sending.
- Organising mail with labels, folders and read status, for example archiving newsletters.
- Sending weekly or team reports using the HTML templates in the skill's assets folder.
- Setting up inbox rules that move or mark incoming mail, with confirmation before each rule change.
Frequently asked questions
Is the lark-mail skill free to use?
The source material does not state pricing. The skill is published under the MIT licence, and it requires the lark-cli tool and a Feishu or Lark account with mail authorisation. Check the lark-cli project and your Feishu plan for any costs.
What do I need before installing lark-mail?
You need the lark-cli command-line tool, a Feishu or Lark account, and mail authorisation through lark-cli auth login --domain mail. For bot identity you also need Feishu developer app permissions.
Does lark-mail work with Gmail or Outlook?
No. The skill only works with Feishu (Lark) mail through lark-cli. It cannot access Gmail, Outlook or other providers.
Will the agent send my emails automatically?
Not by default. Send commands save a draft, and sending needs the --confirm-send flag and your explicit confirmation before the agent runs it. The skill's rules forbid sending without your permission.
Why does the agent ask me to confirm before deleting?
Deletes, trash actions, cancelling scheduled sends and rule changes are treated as sensitive. The agent shows a preview of the affected items, including the count for batch operations, and waits for your approval.
What should I do if the agent cannot access my mailbox?
Check that lark-cli is installed and that you have run lark-cli auth login --domain mail. Without that authorisation the skill cannot access any mailbox. If you use bot identity, confirm the app has the required permissions in the Feishu developer console.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
lark-mail suits people who handle their daily inbox in Feishu or Lark and want an agent to summarise, organise and draft, with a human approving every send. It is a poor fit if you want autonomous sending, or if your mail is on Gmail, Outlook or another provider.
Download the skill