shareAI-lab

Lark Channel — Development skill for Claude Code

Development community

Feishu/Lark group chats connected to Claude Code agents.

How to install Lark Channel

This entry records only its repository, not the path inside it, so there is no exact command to give. Open shareAI-lab/lark-channel and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Lark Channel does

Feishu/Lark group chats connected to Claude Code agents. Per-group isolated sessions, streaming card responses, deep Feishu integration.

Alternatives in Development

Ars Revision — ARS academic-paper revision mode — revised draft + R&R responses 4.5k ★
Claude Token Efficient — One CLAUDE.md file 3.8k ★
Beautiful Feishu Whiteboard — 35 curated colour palette styles for building beautiful, editable Feishu / Lark (飞书) whiteboards 687 ★

README

lark-channel

Connect Feishu/Lark group chats to [Claude Code](https://claude.ai/code) agents. Each group is an isolated workspace: the agent has persistent session context, streams responses as real-time Feishu cards, and its working directory is a dedicated folder on disk.

User: what are the recent open issues?
       |
       v  ~100ms
[EYES reaction]       ← agent received
       |
       v
[text card]           ← streaming, token by token
  Let me check...
       |
       v
[Bash card]           ← tool call
  gh issue list --state open
  ─────────────────────────────
  #142 Fix streaming bug · 2h ago
  #141 Add webhook support · 5h ago
       |
       v
[text card]           ← streaming result
  Found 10 open issues...
       |
       v
[Done · 4s | 2 tool calls]   ← DONE reaction added

The Mental Model

**Group = terminal window.** Each Feishu group is a conversation workspace: one agent, one `cwd`, one persistent session. Asking a follow-up question in the same group picks up where the last message left off — the agent remembers.

**Manager group = terminal manager.** One special group acts as your command center. Tell it what you want to work on; it creates a new group (workspace) with the right agent persona, dedicated folder, and config — then adds you. When you're done, tell it to close the workspace.

Manager group
├── "start a workspace to debug the payment service"
│       → creates group "payment-debug"
│       → creates ~/.lark-channel/workspaces/payment-debug/
│       → writes agents/payment-debug.md (persona: backend debugging agent)
│       → restarts bridge, adds you to group
│
├── "list my workspaces"
│       → shows all active groups + their purposes
│
└── "close payment-debug"
        → archives group, workspace folder stays on disk

Each workspace group has one coding agent running against its folder. Everything the agent does — clones, files, scripts — lands in that folder.