Back to Details

argument-validator

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Argument Validator

论点验证器

Overview

概述

Provide a repeatable workflow for turning informal arguments into formal structure, identifying key assumptions, and checking validity and soundness with optional Lean formalization.
提供一套可重复的工作流，将非形式化论点转化为形式化结构，识别关键假设，并通过可选的Lean形式化来检查有效性和可靠性。

When to Use

使用场景

Use this skill when the user asks to:
  • Validate or critique an argument
  • Formalize an argument in logic or Lean
  • Identify hidden assumptions or missing premises
  • Test an argument with counterexamples
  • Research whether premises are supported by evidence
Do not use this skill for:
  • Simple opinion questions without an argument
  • Purely stylistic rewrites
  • Codebase-only reasoning tasks
当用户提出以下需求时使用本技能：
  • 验证或批判某个论点
  • 用逻辑或Lean对论点进行形式化
  • 识别隐藏假设或缺失的前提
  • 用反例测试论点
  • 研究前提是否有证据支撑
请勿在以下场景使用本技能：
  • 无论点的简单观点类问题
  • 纯风格化改写
  • 仅针对代码库的推理任务

Workflow

工作流

1. Clarify Goal and Scope

1. 明确目标与范围

Ask for missing information before formalization:
  • Request the full argument text, conclusion, and intended audience
  • Ask for definitions of ambiguous terms and the domain of discourse
  • Ask whether the user wants logical validity, empirical soundness, or both
在形式化之前询问缺失的信息：
  • 请求完整的论点文本、结论和目标受众
  • 询问模糊术语的定义和讨论领域
  • 询问用户是需要检查逻辑有效性、实证可靠性，还是两者都需要

2. Extract Argument Structure

2. 提取论点结构

Restate the argument as a numbered list:
  • Separate explicit premises from implicit assumptions
  • Label each premise as logical, definitional, or empirical
  • Note any ambiguous terms or scope shifts
将论点重述为编号列表：
  • 区分明确前提与隐含假设
  • 将每个前提标记为逻辑型、定义型或实证型
  • 记录模糊术语或范围转移情况

3. Formalize the Logic

3. 逻辑形式化

Translate into a precise formal representation:
  • Choose the smallest logic that fits (propositional or first-order)
  • Define symbols and predicates explicitly
  • Encode the argument as premises → conclusion
  • Flag quantifier order and scope changes
将论点转化为精确的形式化表示：
  • 选择最贴合的最小逻辑系统（命题逻辑或一阶逻辑）
  • 明确定义符号和谓词
  • 将论点编码为「前提→结论」的形式
  • 标记量词顺序和范围变化

4. Check Logical Validity

4. 检查逻辑有效性

Attempt a derivation from premises to conclusion:
  • Identify the first point where the proof fails
  • Produce the minimal additional assumption needed for validity
  • Provide a counterexample model when possible
尝试从前提推导结论：
  • 识别证明失败的首个节点
  • 给出使论证有效的最小附加假设
  • 尽可能提供反例模型

5. Formalize in Lean (Optional)

5. 用Lean进行形式化（可选）

When the user wants machine-checking, offload to a Lean formalizer:
  • Check for Lean availability (
    lean --version
    or 
    ~/.elan/bin/lean --version
    )
  • Use 
    lean --stdin
    for quick checks when no project exists
  • Ask the formalizer to return a compilable Lean snippet plus missing lemmas
当用户需要机器验证时，将任务转交给Lean形式化工具：
  • 检查Lean的可用性（
    lean --version
    或 
    ~/.elan/bin/lean --version
    ）
  • 当无项目时，使用 
    lean --stdin
    进行快速检查
  • 要求形式化工具返回可编译的Lean代码片段以及缺失的引理

6. Validate Assumptions with Research Agents

6. 借助研究Agent验证假设

For each empirical or contestable assumption:
  • Spawn one research subagent per assumption
  • Provide the exact assumption and desired standard of evidence
  • Require a summary, sources, and a confidence rating
  • Run subagents in parallel when there are multiple assumptions
针对每个实证型或有争议的假设：
  • 为每个假设生成一个研究子Agent
  • 提供具体假设和所需的证据标准
  • 要求返回摘要、来源和置信度评级
  • 当有多个假设时，并行运行子Agent

7. Synthesize the Final Analysis

7. 合成最终分析报告

Deliver a structured summary:
  • Validity verdict (valid / invalid) with justification
  • Soundness verdict (supported / unsupported / unknown)
  • List of key assumptions and their evidence status
  • Suggested revisions that would strengthen the argument
交付结构化的摘要：
  • 有效性判定（有效/无效）及理由
  • 可靠性判定（有支撑/无支撑/未知）
  • 关键假设列表及其证据状态
  • 可增强论点的修订建议

Subagent Prompts

子Agent提示词

Formalizer Agent (Logic + Lean)

形式化Agent（逻辑+Lean）

Use a 
general
subagent to formalize the argument:
You are a FORMALIZER agent.

INPUT:
- Argument text
- Extracted premises + conclusion
- Draft formalization (symbols and formulas)

TASK:
1. Tighten the formalization (minimal logic).
2. Identify missing premises or implicit assumptions.
3. Attempt a Lean formalization.
4. If proof fails, explain where and why.

OUTPUT:
- Refined formalization
- Lean theorem statement
- Lean proof sketch or error explanation
- List of missing assumptions
使用「general」子Agent来形式化论点：
You are a FORMALIZER agent.

INPUT:
- Argument text
- Extracted premises + conclusion
- Draft formalization (symbols and formulas)

TASK:
1. Tighten the formalization (minimal logic).
2. Identify missing premises or implicit assumptions.
3. Attempt a Lean formalization.
4. If proof fails, explain where and why.

OUTPUT:
- Refined formalization
- Lean theorem statement
- Lean proof sketch or error explanation
- List of missing assumptions

Assumption Research Agents

假设研究Agent

Use one 
general
subagent per assumption:
You are a RESEARCHER agent.

ASSUMPTION:
[insert assumption]

TASK:
1. Use available web tools to find supporting or refuting sources.
2. Summarize evidence with citations.
3. Rate confidence (low/medium/high).

OUTPUT:
- Evidence summary
- Source list with URLs
- Confidence rating
- Notes on conflicts or gaps
为每个假设使用一个「general」子Agent：
You are a RESEARCHER agent.

ASSUMPTION:
[insert assumption]

TASK:
1. Use available web tools to find supporting or refuting sources.
2. Summarize evidence with citations.
3. Rate confidence (low/medium/high).

OUTPUT:
- Evidence summary
- Source list with URLs
- Confidence rating
- Notes on conflicts or gaps

Output Format

输出格式

Provide results in this order:
  1. Restated argument (premises → conclusion)
  2. Formalization (symbols + formulas)
  3. Validity analysis (proof gap or confirmation)
  4. Lean check results (if performed)
  5. Assumptions table (premise, type, evidence, status)
  6. Recommendations or questions to resolve uncertainty
按以下顺序提供结果：
  1. 重述论点（前提→结论）
  2. 形式化表示（符号+公式）
  3. 有效性分析（证明缺口或验证结果）
  4. Lean检查结果（若已执行）
  5. 假设表格（前提、类型、证据、状态）
  6. 建议或用于解决不确定性的问题