feishu-agent-add

This skill is the conversational front end for the local script scripts/add_feishu_agent.py.

This project is designed for OpenClaw users, but the skill name intentionally stays short: feishu-agent-add.

When To Use

Use this skill when the user wants to:

• add a new Feishu-connected OpenClaw agent
• avoid hand-editing openclaw.json
• configure a new agent through a few follow-up questions
• get a ready-to-run one-line command for advanced usage

Core Rule

Do not hand-edit openclaw.json unless the user explicitly asks for manual fallback.

Prefer running:

python3 scripts/add_feishu_agent.py ...

The script is the execution core. This skill should mainly:

1. understand the user's request
2. ask only for missing required fields
3. preview the plan
4. run the script
5. summarize the result and next steps

Required Inputs

Collect these fields before execution:

• agent_name
• purpose
• agent_id
  • if missing, propose one derived from the name
• app_id
• app_secret

These can use defaults unless the user says otherwise:

• workspace_path
  • default: ~/.openclaw/workspace-{agent_id}
• model
  • default: inherit from the current OpenClaw config
• enable_agent_to_agent
  • default: true
• workspace_mode
  • default: auto
• init_templates
  • default: true

Conversational Flow

1. Parse what the user already gave

For a request like:

帮我增加一个名字叫小红书运营，用来做内容选题和文案生成的飞书agent

extract:

• agent_name = 小红书运营
• purpose = 内容选题和文案生成
• agent_id = xiaohongshu or another short slug candidate

If agent_id is missing, propose one instead of asking an open-ended question.

2. Ask the minimum follow-up questions

Only ask for the missing required fields. Prefer one compact message.

Typical follow-up:

• proposed agent_id
• Feishu App ID
• Feishu App Secret

Only ask about optional fields if the user indicates they care.

3. Preview before execution

Before running the script, summarize:

• agent name
• agent id
• purpose
• workspace path
• whether agent-to-agent collaboration will be enabled

4. Run the script

Run from the skill directory:

python3 scripts/add_feishu_agent.py \
  --agent-id <agent-id> \
  --agent-name "<agent-name>" \
  --purpose "<purpose>" \
  --app-id <app-id> \
  --app-secret <app-secret> \
  --json-output \
  --yes

Add optional flags only when needed:

• --model <model>
• --workspace-path <path>
• --disable-agent-to-agent
• --workspace-mode cli|mkdir|auto
• --no-init-templates
• --dry-run

Advanced User Mode

If the user prefers a single terminal command, give them a ready-to-run example instead of a manual JSON recipe.

Use this pattern:

python3 scripts/add_feishu_agent.py \
  --agent-id trader \
  --agent-name "交易小助手" \
  --purpose "股票和 ETF 分析" \
  --app-id cli_xxx \
  --app-secret secret_xxx \
  --yes

Output Expectations

After execution, summarize:

• whether config was written successfully
• the workspace path
• whether starter files were initialized
• that OpenClaw should be restarted
• where to refine the agent identity, usually SOUL.md

If the script fails, report the concrete reason and do not improvise partial manual edits unless the user asks for that fallback.

Notes

• The script already handles validation, backup, and config updates.
• Prefer --dry-run first when the user asks for a preview.
• If the user asks how to install or use this project, point them to README.md.