Taixu Debate logo

Taixu Debate

Community
fainshare
taixu-debate

五角色四轮混合盲辩压测:奥斯本/黑格尔/贝叶斯/费曼/德博诺(发散/收敛/验证/审查/归真) 对高 stakes 命题(投资决策/战略选择/认知求真)对撞辩论,产出辩论精华 + 三区(共识/分歧/未决)+ 综合判断。 触发词:辩论、辩论一下、太虚辩论、混合盲辩、taixu-debate、/taixu-debate。 NOT for: 日常分析、pipeline 替代、低 stakes 命题、不可证伪的泛论。

Overview

Publisherfainshare
RepositoryLiam-Skills
Skill nametaixu-debate
Stars
113
Forks
19
Bundled files
12
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 12 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by fainshare on GitHub. Read the source before you install it.

Installation

Install the Taixu Debate AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/fainshare/Liam-Skills/tree/main/skills/taixu-debate
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/fainshare/Liam-Skills.git /tmp/Liam-Skills
mkdir -p .claude/skills
cp -r /tmp/Liam-Skills/skills/taixu-debate .claude/skills/taixu-debate
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Taixu Debate in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Taixu Debate on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Taixu Debate is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Taixu-Debate 太虚辩论

五个认知角色对同一命题盲辩交锋,在隔离中独立立论,在汇总后交叉反驳。 不是串联审查(太虚五转 pipeline),是四轮对撞压测。

概述

高 stakes 命题定论前(投资决策、战略选择、认知求真),太虚五转提供串联审查范式(pipeline 顺序执行),但缺少五角色四轮对撞的压测机制。taixu-debate 让五位大师人格——奥斯本(发散)、黑格尔(收敛)、贝叶斯(验证)、费曼(审查)、德博诺(归真)——在同一辩题上同时盲立论,交叉回应后防御/修正立场,产出辩论精华 + 三区(共识/分歧/未决)+ 综合判断 + 按命题类型分流的结论产出。

角色不是五原子 skill 的流程复刻,而是从五原子认知功能提炼的立场化辩论人格:各自带着自己的核心关切和攻击面进入辩论,互相挑战对方的盲区。五原子 skill 零修改,角色快照是独立副本。辩论发言、引用标记、报告中一律使用大师名;英文 ID(divergent 等)仅用于 subagent 类型与文件命名。

五位大师之外设主持人格西塞罗(Cicero,罗马议事主持人、交叉质证律师、文辞家):编排层的人格化,负责赛程控制、引用校验、数据审查、三区归纳、裁决书与实录文书。西塞罗由主 agent 扮演,有程序权与记录权但没有裁决权,永不下场辩论。快照见 references/roles/host.md。

场景限定(R6)

触发条件(pipeline 后置):

  • 命题必须是高 stakes 判断,属三类之一:投资决策型(有具体标的与仓位动作)、选择决策型(技术选型/战略路径/方案取舍)、认知求真型(理解现象、检验叙事)
  • 已经过太虚五转 pipeline 至少一轮收敛(不允许对原始发散结果直接辩论)
  • 命题可证伪(含明确的方向性预测,非纯描述性陈述)

假命题入场识别(拒绝以下输入):

  • 不可证伪的陈述(如"市场是有效的")
  • 纯历史回顾无前瞻预测
  • 不可证伪的泛论(如"AI 会改变一切");认知求真型辩题允许无具体标的,但须含可检验的子命题
  • 已形成强共识且无分歧空间的命题

三层编排架构

层级机制环境隔离保证
首选workflow.js 机器级编排Claude Code execution / Qoder CLI dynamic workflow API运行时隔离
降级主 agent 用 Agent 工具 dispatch 自定义 subagentQoder IDE / 无 workflow 环境平台上下文隔离
最后手段主 agent 同上下文扮演五角色任何环境无隔离(须标注)

降级标注规则:报告 frontmatter 必须含 execution_mode: workflow | subagent_dispatch | inline_roleplay。使用最后手段时,报告正文顶部须显式标注"⚠️ 同上下文扮演降级模式——角色隔离未保证,交锋质量可能受损"。

降级路径执行协议

workflow.js 可用时(Claude Code execution / Qoder CLI dynamic workflow): 西塞罗(主 agent 扮演)按 scripts/workflow.js 定义的 10 phases 自动执行(EvidenceBase → ConceptAlignment → BlindPosition → Aggregate → CrossResponse → Aggregate → Defense → Aggregate → ClosingCross → Report),隔离由运行时保证。

workflow.js 不可用时(Qoder IDE / 无 workflow 环境): 主 agent 按 references/debate-protocol.md 的"降级路径执行方式"用 Agent 工具 dispatch 自定义 subagent。关键规则:

  • R1/R2/R3/R4 每轮用一条消息并行 dispatch 多个 Agent 调用(实现 fan-out)
  • 他方立论给原文不给摘要
  • 引用机械校验由主 agent 在报告生成前执行
  • 沉默角色在后续轮次跳过;R4 默认沉默制——仅开放战线非空的角色 dispatch

最后手段(无 subagent 能力): 主 agent 同上下文扮演五角色,报告 frontmatter execution_mode: inline_roleplay,正文顶部须标注降级警告。此模式角色隔离未保证,交锋质量可能受损。

双产出 + 自动润色(所有执行模式): 无论哪种执行模式,辩论结束后主 agent 产出两份文档——摘要版裁决书(≤1200 字,含辩论精华 + 三区 + 综合判断 + 结论产出)+ 全程实录汇总稿(辩论赛标准格式,含立场演化表 + 数据可靠性审查),然后自动调用 liam-polish skill 对两份文档做润色(消除 AI 味与翻译腔),全程不向用户提问。摘要版裁决书必须面向人读(禁止术语黑话,鼓励类比和具象化,允许结构化表格,普通人能直接读懂)。详见 references/debate-protocol.md 的「报告生成」节与 templates/report-template.md 的「双产出定义」「自动润色流程」「面向人读(硬约束)」「报告结构」节。

交付物归档(所有执行模式,R15 硬约束): 双产出润色完成后,以润色稿为正本归档至 docs/taixu-debate-reports/(与 docs/hegel-reports 等五原子 skill 报告目录惯例一致),命名 <topic_slug>-taixu-debate-report-<date>.md / <topic_slug>-taixu-debate-transcript-<date>.md。topic_slug 为英文小写短横线 3-6 词,由西塞罗开辩时从辩题提炼(如 ai-semiconductor-capex-bubble)并写入 metadata.json。归档只留润色后正本,不留 -原文/-润色稿 双轨;过程文件写入本场运行目录 runtime/debate/<date>-<topic_slug>-<run_id>/(run_id 为开辩时刻 HHmmss,开辩时确定、全程不变;同一日期、同一话题的并行辩论也不得共用目录),仅保留 R1-R4 发言原文、metadata.json 等过程文件,交付物一律不留临时区。

辩论协议概述

四轮结构(详见 references/debate-protocol.md):

  1. Phase 0 证据底座:西塞罗向用户询问研究预算(轻量/标准/深度),按命题类型调用研究工具,产出共享 EvidencePack 注入所有 R1 task prompt。投资决策型研究失败则 hard_stop;其他类型软降级并标注"⚠️ 无外部证据底座" 0.5. Phase 0.5 概念对齐(v0.8.0 新增):西塞罗从 EvidencePack 提取 3-5 个核心概念并统一定义(口径边界),写入 <run_dir>/concept_alignment.md,与 EvidencePack 并列注入所有 R1 task prompt。主 agent 直接从 EvidencePack 提炼,不额外调用 LLM(仅 EvidencePack 为空时 1 次)。角色须使用统一定义立论;汇总 #1 时西塞罗事后检查概念使用偏差
  2. R1 盲立论(parallel fan-out):5 角色各自独立对辩题立论,互相不可见
  3. 西塞罗汇总:将 5 份立论合并为公开记录
  4. R2 交叉回应(parallel fan-out):每角色收到自己立论回放 + 他方立论原文,接力反驳须引用具体论点。贝叶斯可在此轮发起一次补充检索(≤3 次 WebSearch),结果追加公开记录供 R3 使用
  5. 西塞罗汇总 #2:将 R2 回应合并为公开记录;标注全场盲区(只标注不引导,注入 R3 task prompt——R4 终辩禁开新战线,R3 是最后一个能承接新维度的发言轮)
  6. R3 防御/修正(parallel fan-out):每角色收到自己 R1+R2 回放 + 他方 R2 原文,防御自身被攻击的论点或修正立场,须引用具体论点
  7. 西塞罗汇总 #3:合并 R3 为公开记录;为每个角色编译开放战线清单(他方 R3 的新攻击/新论证/修正后立场中与该角色相关的条目)
  8. R4 检验交锋与终辩声明(默认沉默制):仅开放战线非空的角色发言。前半段「检验交锋」逐条回应开放战线(只引用 R3);后半段「终辩声明」给出终态立场 + 让步清单,禁开新战线。开放战线全空则跳过 R4 直接出报告

发言五要素:立场、核心论据、对他方论点的反驳(含引用)、沉默声明(无反驳则显式沉默)、本轮小结(≤60字白话,坚持/松动/转向了什么,因为什么)。R4 采用两段式格式(检验交锋 + 终辩声明),详见协议

报告四组件(详见 templates/report-template.md):

  • 辩论精华:2-3 条穿透性引文
  • 三区:共识区 / 分歧区 / 未决区
  • 综合判断:西塞罗补充维度
  • 结论产出:按命题类型分流(投资决策型给行动建议,选择决策型给结论与适用边界,认知求真型给采纳边界与验证跟踪)

角色注册表

5 个辩手角色 + 1 个主持人格定义见 references/role-registry.md。角色快照见 references/roles/。

角色来源核心关切
奥斯本(发散 · divergent)osborn可能性空间是否充分打开?
黑格尔(收敛 · convergent)hegel不确定项是否收敛?
贝叶斯(验证 · verifier)bayes结论经得起外部数据打脸吗?
费曼(审查 · auditor)feynman你在骗自己吗?
德博诺(归真 · integrator)debono被否掉的东西有没有被忽略的价值?
西塞罗(主持 · host)编排层人格化程序与实录是否守住?(非辩手,无 subagent 定义)

token 硬上限(R8)

单场全流程 token 消耗 ≤ 4.5× 单一强 prompt 基线(随研究预算档位浮动:轻量 4.3×、标准 4.5×、深度 5.0×)。超限硬终止。具体口径见 references/debate-protocol.md。

验证实验

skill 投入使用前须经等 token 盲评实验验证增量价值。实验协议见 references/validation-experiment.md。

不做的事

  • 不替代太虚五转 pipeline(互补关系,非替代)
  • 不做动态轮次判定——轮次数恒定为 4 轮,发言面动态由沉默机制与 R4 默认沉默制内生实现(多轮迭代辩论 v2 deferred)
  • 不修改五原子 skill 的任何文件
  • 不做对抗烈度控制(v2 deferred)

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Taixu Debate AI skill do?

五角色四轮混合盲辩压测:奥斯本/黑格尔/贝叶斯/费曼/德博诺(发散/收敛/验证/审查/归真) 对高 stakes 命题(投资决策/战略选择/认知求真)对撞辩论,产出辩论精华 + 三区(共识/分歧/未决)+ 综合判断。 触发词:辩论、辩论一下、太虚辩论、混合盲辩、taixu-debate、/taixu-debate。 NOT for: 日常分析、pipeline 替代、低 stakes 命题、不可证伪的泛论。

Why use Taixu Debate on TypingMind?

Because you install it once and use it with any model. Taixu Debate is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Taixu Debate in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/fainshare/Liam-Skills/tree/main/skills/taixu-debate. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Taixu Debate?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Taixu Debate?

As many as you like. As long as a model supports skills, you can use Taixu Debate with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Taixu Debate AI skill free?

It is published on GitHub by fainshare. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇