← Back to Skills Marketplace
Agent Factory (Walter)
by
beyondbright
· GitHub ↗
· v2.1.1
· MIT-0
178
Downloads
0
Stars
1
Active Installs
10
Versions
Install in OpenClaw
/install walter-agent-factory
Description
创建新的 OpenClaw Agent 并自动配置飞书机器人。当用户说"帮我创建 Agent"、"新建机器人"、"添加新 agent"、"配置新机器人"时触发。用户提供：agent 名称、飞书 appId、appSecret、角色定位。执行完毕后汇报结果。
Usage Guidance
This skill will run commands that modify your OpenClaw installation and persist the provided Feishu appSecret into ~/.openclaw/openclaw.json, then restart the gateway. Before installing or running it: 1) Confirm you trust the author; 2) Manually review the SKILL.md steps; 3) Ensure openclaw, clawhub, curl/jq (or PowerShell) are installed and understand the commands will run with your user privileges; 4) Back up ~/.openclaw/openclaw.json and verify file permissions (secrets are stored in plain JSON); 5) Prefer to run the script interactively yourself rather than granting an automated agent permission to execute it, or at minimum run in a safe/non‑production environment; 6) If you plan to install a third‑party skill via clawhub, verify the skill name/source first. The metadata should be updated to declare required binaries and the sensitive nature of persisted credentials — treat the current omission as a red flag.
Capability Analysis
Type: OpenClaw Skill
Name: walter-agent-factory
Version: 2.1.1
The skill acts as an administrative tool to automate the creation of new agents, but it performs high-risk operations including modifying the core 'openclaw.json' configuration and handling sensitive Feishu 'appSecret' credentials. It is vulnerable to command injection in both the PowerShell and Bash scripts (SKILL.md) because it uses 'Invoke-Expression' and unsanitized variables to execute CLI commands. Additionally, it copies 'MEMORY.md' from the current workspace to the new one, which could lead to unintentional data leakage of sensitive conversation history between agents.
Capability Assessment
Purpose & Capability
The description (create OpenClaw agent and configure Feishu/Lark bot) aligns with the SKILL.md steps: validate Feishu appId/appSecret, create agent, add account and bindings in ~/.openclaw/openclaw.json, create workspace files, optionally install a clawhub skill, enable channels/tools and restart the gateway. However, the skill metadata declares no required binaries or env vars while the instructions clearly assume presence of openclaw, clawhub, curl/jq (or PowerShell) and access to HOME/USERPROFILE and ~/.openclaw — this mismatch should be corrected or explained by the author.
Instruction Scope
The runtime instructions read and write the user's OpenClaw global config (~/.openclaw/openclaw.json), add persistent Feishu credentials (appSecret) to it, create files in a new agent workspace, install third‑party skills, toggle global channel/tool flags, and restart the gateway. Those operations are coherent with the stated purpose but are high‑impact: they modify global config, persist secrets in plain JSON, and affect other agents via gateway restart.
Install Mechanism
This is instruction‑only (no install spec), so nothing is written by the skill package itself. But the instructions assume external tooling (openclaw, clawhub, curl, jq or PowerShell). The absence of declared required binaries/tools in metadata is an omission that makes automated safety checks unreliable.
Credentials
No environment variables or primary credential are declared in the registry metadata, yet the flow requires the user to supply sensitive Feishu credentials (appId/appSecret) which will be written to ~/.openclaw/openclaw.json. The skill also expects access to HOME/USERPROFILE and may use OPENCLAW_DIR if set. Persisting appSecret in an on‑disk JSON config without discussing encryption or file protections is sensitive and should be justified/mitigated.
Persistence & Privilege
The skill modifies global OpenClaw configuration, adds channel accounts/bindings, enables channel tools, and restarts the gateway — all of which have system‑wide effects on other agents and users. The skill is not 'always:true', but its instructions grant it substantial lasting impact on the host environment.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install walter-agent-factory - After installation, invoke the skill by name or use
/walter-agent-factory - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.1
更新 soul 示例值
v2.1.0
Add verification steps after each critical operation; fix MEMORY.md content check to detect empty templates
v2.0.0
Fix Windows path issues: replace ~ with C:\Users\beyon, remove unused variables
v1.3.0
新增 MEMORY.md 复制逻辑，新 agent 创建时自动带入永久记忆
v1.2.0
新增 skill 参数，skill 安装到新 agent workspace；Step 8 开启飞书工具权限；修复 Step 7 跳步问题
v1.1.0
优化：原子写入、Depth 100、路由冲突检查；Step 7 明确 skill 安装到 active workspace
v1.0.3
新增 Linux/macOS Bash 脚本版本，支持跨平台使用
v1.0.2
将 create-agent.ps1 完整脚本合并至 SKILL.md，确保完整发布
v1.0.1
补充 scripts/create-agent.ps1 脚本文件
v1.0.0
Initial release: 自动创建新 OpenClaw Agent 并绑定飞书机器人
Metadata
Frequently Asked Questions
What is Agent Factory (Walter)?
创建新的 OpenClaw Agent 并自动配置飞书机器人。当用户说"帮我创建 Agent"、"新建机器人"、"添加新 agent"、"配置新机器人"时触发。用户提供：agent 名称、飞书 appId、appSecret、角色定位。执行完毕后汇报结果。 It is an AI Agent Skill for Claude Code / OpenClaw, with 178 downloads so far.
How do I install Agent Factory (Walter)?
Run "/install walter-agent-factory" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Agent Factory (Walter) free?
Yes, Agent Factory (Walter) is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Agent Factory (Walter) support?
Agent Factory (Walter) is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Agent Factory (Walter)?
It is built and maintained by beyondbright (@beyondbright); the current version is v2.1.1.
More Skills