OpenClaw for Feishu Teams: 3 Workflows That Work

What actually works in Feishu team scenarios

The best first workflows for Feishu teams are not the flashiest automations. They are the ones that produce visible, reviewable, and rollbackable output.

Focus on these three:

1. Group messages → daily reports
2. Meeting minutes automation
3. Link collection → knowledge base

Each delivers tangible output that teams can verify before trusting more automation.

Workflow 1: Group messages → daily reports

Let the bot summarize key discussions, decisions, and action items from group chat and write them into a Feishu document.

Prerequisites

openclaw plugins install @openclaw/feishu
openclaw skills install @clawhub/summarize

AGENTS.md example

# Team Daily Report Agent

identity: You are a team assistant for Feishu. You do not reply to group messages directly. You only organize and write.

workflow:
  - At 18:00 daily, collect the day's key discussions, resolutions, and todos from the designated group.
  - Output format: themed daily report sections, written to the specified Feishu doc.
  - Do not send messages to the group. Only read and write documents.

permissions:
  - Read group messages
  - Write to specified Feishu document only
  - No send_as_user

Skill recommendation

Install @clawhub/summarize for summarization. Add @clawhub/file-manager if you need structured file output.

Workflow 2: Meeting minutes automation

After a meeting, turn recording or transcript into structured minutes with summaries and action items.

Skills needed

openclaw skills install @clawhub/summarize
openclaw skills install @clawhub/file-manager

AGENTS.md example

# Meeting Minutes Agent

identity: You process meeting recordings and transcripts into structured minutes.

workflow:
  - Accept meeting transcript or recording reference.
  - Extract: key points, decisions, action items with owners.
  - Output: structured minutes in specified Feishu doc format.
  - Do not modify other documents or send messages.

permissions:
  - Read transcript/recording
  - Write to designated minutes document only

Teams should always review before publishing. Do not auto-post to broad channels without human approval.

Workflow 3: Link collection → knowledge base

Automatically capture links shared in group chat and add them to a Feishu knowledge base or multidimensional table with title, summary, and category.

Skills needed

openclaw skills install @clawhub/web-search

Configure Content Collector to grab links from group chat, fetch metadata, and write to your knowledge base.

AGENTS.md example

# Knowledge Base Collector Agent

identity: You curate shared links into the team knowledge base.

workflow:
  - Detect links shared in designated groups.
  - Fetch title, summary, and suggest category.
  - Write entry to Feishu knowledge base or multidimensional table.
  - Do not post messages or modify other content.

permissions:
  - Read group messages for links
  - Write to knowledge base / table only
  - No send_as_user

Permission control: stricter for teams

Team scenarios magnify permission risks. Apply these rules:

• Least privilege: Only grant permissions needed for the current workflow.
• Avoid send_as_user: Unless explicitly required, do not let the bot send messages as a user.
• Document write scope: Grant write access to specific documents, not global document write.

Check your setup:

openclaw security audit --deep
openclaw config list | grep -E "auth|permission"

Multi-agent division of labor

Split responsibilities across agents instead of one super-agent:

Agent 1: Info Organizer
  - Read group messages only
  - Write to specified documents only
  - No reply capability

Agent 2: Task Manager
  - Read todos from group messages
  - Write to task list
  - Send one daily task summary

Agent 3: Knowledge Curator
  - Collect links from chat
  - Write to knowledge base
  - No access to other data

Each agent has its own workspace and permissions. A problem in one does not cascade.

Pre-launch checklist

Before going live, confirm:

• Bot reply loop works (mention → response)
• Document writes succeed and format is correct
• Permissions match the minimum required
• Team knows what the bot can and cannot do
• There is a clear human review step
• openclaw security audit --deep passes

Next steps

• Feishu channel setup
• Feishu initialization guide
• Feishu security governance
• Multi-agent architectures
• Official docs → docs.openclaw.ai