← 返回 Skills 市场
78
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install virtual-forum
功能描述
虚拟论坛：让蒸馏的人物Skill就特定话题展开讨论。 v5.0使用Claude Code CLI实现真正的多agent并行辩论。 内置博弈论分析模块（信号博弈、议价博弈、联盟博弈、行为经济学）。 触发词：「虚拟论坛」「发起讨论」「圆桌会议」「辩论」「主持讨论」「让XX YY讨论」
安全使用建议
This skill is coherent with its purpose (it needs persona text to simulate participants) but it will read local SKILL.md files and send their contents to an external Claude service via the Claude Code CLI. Before installing or running it: 1) Do not keep secrets, API keys, or credentials inside any SKILL.md or untrusted files in your skills directory; sanitize or remove sensitive data. 2) Inspect v5/debate_parallel.sh and shared loadSkill() to confirm which paths are read and how prompts are constructed. 3) Be aware you must install and authenticate the external 'claude' CLI (which uses its own API key) — the skill does not declare that credential requirement in the registry metadata. 4) If you want to test safely, run the skill in an isolated environment or container, point SKILLS_DIR to a sanitized copy containing only persona texts, and set OUTPUT_DIR to a controlled folder. 5) If you cannot guarantee SKILL.md files are free of secrets, avoid running the parallel‑Claude mode or modify the script to avoid transmitting raw skill files (e.g., extract only allowed fields or use local-only simulation).
功能分析
Type: OpenClaw Skill
Name: virtual-forum
Version: 5.0.3
The skill implements a 'Virtual Forum' that intentionally reads the full content of other skills from the user's local directory (~/.openclaw/workspace/skills) and transmits them to an external service via the 'claude' CLI, as documented in the security warning in SKILL.md and implemented in v5/debate_parallel.sh. While this is presented as a feature for multi-agent debates, it facilitates the exfiltration of potentially sensitive data (such as mental models or API keys) stored in other skills. Furthermore, CHANGELOG.md explicitly states that the v5.0 architecture is designed to 'completely bypass OpenClaw sessions limits' by using external shell scripts, which is a technique used to evade platform-imposed security and resource constraints.
能力标签
能力评估
Purpose & Capability
The name/description (multi‑agent virtual forum that uses distilled persona Skills) matches the code and scripts: it loads other Skills (SKILL.md) and composes system prompts to drive Claude Code processes. That behavior is coherent with the stated purpose. Minor inconsistency: the package/registry metadata declares no required env vars or binaries, but the README/SKILL.md and scripts require the Claude Code CLI and configured API auth.
Instruction Scope
SKILL.md and v5/debate_parallel.sh explicitly read local SKILL.md files (e.g., ~/.openclaw/.../SKILL.md), inject their contents into system prompts, and send them to an external model via the Claude CLI. That means arbitrary local Skill content (and anything embedded in those files) will be transmitted to an external service. The SKILL.md does warn about this, but the runtime instructions give the agent broad discretion to load and forward local skill files and to construct system prompts — this is beyond simple internal computation and is sensitive by nature.
Install Mechanism
No install spec is included (instruction/code bundle only), which reduces supply‑chain risk. However the README instructs users to install an external Claude Code CLI (brew or npm) and to run debate_parallel.sh which will spawn external processes. There are no downloads from untrusted URLs in the bundle itself. The missing formal declaration of required external CLI in the registry metadata is a practical shortcoming.
Credentials
The skill asks users (in docs) to set SKILLS_DIR and OUTPUT_DIR and requires a configured Claude CLI/API key, but the registry metadata lists no required env vars/credentials. The skill will read arbitrary files under the user's skills directory (SKILL.md of other persona Skills) which may contain secrets or API keys if users put them there. While reading persona Skill files is functionally necessary, transmitting them to an external API is high‑sensitivity and should be explicitly declared and minimized.
Persistence & Privilege
The skill does not request always:true and does not modify other skills' configurations according to the provided files. It writes outputs to an output directory and uses temporary files; nothing in the manifest suggests it gains elevated or permanent platform privileges.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install virtual-forum - 安装完成后，直接呼叫该 Skill 的名称或使用
/virtual-forum触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v5.0.3
v5.0.3: 重组目录，博弈论移至v5/game-theory/
v5.0.2
安全修复：移除硬编码路径，添加安全警告
v5.0.1
v5.0.1: 修复语法错误和增强错误处理
元数据
常见问题
虚拟论坛 是什么？
虚拟论坛：让蒸馏的人物Skill就特定话题展开讨论。 v5.0使用Claude Code CLI实现真正的多agent并行辩论。 内置博弈论分析模块（信号博弈、议价博弈、联盟博弈、行为经济学）。 触发词：「虚拟论坛」「发起讨论」「圆桌会议」「辩论」「主持讨论」「让XX YY讨论」. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 78 次。
如何安装 虚拟论坛？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install virtual-forum」即可一键安装，无需额外配置。
虚拟论坛 是免费的吗？
是的，虚拟论坛 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
虚拟论坛 支持哪些平台？
虚拟论坛 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 虚拟论坛？
由 Yirong（@erongcao）开发并维护，当前版本 v5.0.3。
推荐 Skills