Ai Court Skill

以明朝内阁制为蓝本的多 Agent 协作系统 - 一键部署你的 AI 朝廷

Security Scan

The skill's files, instructions, and configuration templates are coherent with a multi‑agent orchestration framework (AI Court) and largely match the stated purpose, but review a small install script and any credentials/templates before use.

Purpose & Capability

Name/description (multi‑agent 'court' orchestration) aligns with the included agent configs, SOUL templates, and openclaw.json provider/channel placeholders; the files are what you'd expect for a multi‑agent orchestration skill.

Instruction Scope

SKILL.md gives simple install/copy/start steps that stay within purpose, but many agent identity files reference runtime commands (sessions_send, message --channel discord, node scripts/task-store.js), GitHub webhook triggers, and Discord/GitHub integration. These are coherent for collaboration but mean the skill expects you to enable external channels and webhooks — review what you enable and whether you want those integrations active.

Install Mechanism

There is no formal install spec (instruction‑only), which is lower risk. However a scripts/post-install.sh file is present (1061 bytes). Its contents were not shown in the provided dump — any post‑install script run by the platform could execute commands on install, so inspect it before running or allow the install tool to show/confirm actions.

Credentials

The skill does not declare required env vars, but the templates clearly expect LLM provider API keys and optional Discord bot tokens / webhook config. Requesting those credentials is proportionate for this use case, but they are sensitive — do not paste long‑lived tokens into files that may be committed; prefer environment variables and limit token scopes.

Persistence & Privilege

The skill does not set always:true and follows normal agent configuration patterns (workspaces under $HOME, sandbox modes). Some agents have sandbox mode toggled off and subagents allowed; this is expected for a coordination framework but increases what agents can do if you run them with full privileges — review sandbox settings and channel/gateway configs before enabling.

Assessment

This package appears to be a coherent multi‑agent orchestration template. Before installing: 1) Backup ~/.openclaw (or your OpenClaw config) — the recommended copy commands will write many files into your home config. 2) Inspect scripts/post-install.sh and any JS scripts referenced (e.g., task-store.js) to see what they will run; do not run unknown install scripts without review. 3) Do not hardcode API keys, Discord bot tokens, or GitHub tokens into the copied openclaw.json; prefer environment variables and minimal scopes. 4) If you enable Discord/webhook/GitHub integrations, expect agents to send/receive messages and to be triggered by webhooks — only provide tokens to bots/accounts you control and limit permissions. 5) Check sandbox settings in the copied openclaw.json (some agents use sandbox: off/all) and tighten them if you want to restrict agent capabilities. If you want, provide the contents of scripts/post-install.sh and any missing runtime scripts (task-store.js) and I can review them for any risky behaviors.

AI Court | 当皇上

一行命令起王朝，三省六部皆 AI。

安装

clawdhub install ai-court

使用

安装后复制配置：

# 明朝内阁制
cd ~/.openclaw && cp -r clawd/skills/ai-court/configs/ming-neige/* .

# 或唐朝三省制
cd ~/.openclaw && cp -r clawd/skills/ai-court/configs/tang-sansheng/* .

# 或现代企业制
cd ~/.openclaw && cp -r clawd/skills/ai-court/configs/modern-ceo/* .

配置 API Key 后启动：

openclaw start

文档

查看 references/ 目录中的详细文档。

链接

GitHub: https://github.com/wanikua/ai-court-skill