erdos-prove — for Claude Code erdos-prove, erdos-banger, community, for Claude Code, ide skills, combinatorics, erdos-problems, formal-verification, mathematics, number-theory

v1.0.0

关于此技能

适用场景: Ideal for AI agents that need erdos problem proving workflow. 本地化技能摘要: This workflow helps you prove Erdos problems in Lean 4 using your Claude Code/Codex subscription instead of paid API calls. It covers cli, combinatorics, erdos-problems workflows.

功能特性

Erdos Problem Proving Workflow
Target Problem: $ARGUMENTS
Tip: If no problem ID provided, check CANDIDATES.md for the current focus problem (see Decision
┌─────────────────────────────────────────────────────────────────┐
│ COST-FREE PROVING WORKFLOW │

# 核心主题

The-Obstacle-Is-The-Way The-Obstacle-Is-The-Way
[4]
[2]
更新于: 2/1/2026

Killer-Skills Review

Decision support comes first. Repository text comes second.

Reference-Only Page Review Score: 10/11

This page remains useful for teams, but Killer-Skills treats it as reference material instead of a primary organic landing page.

Original recommendation layer Concrete use-case guidance Explicit limitations and caution Quality floor passed for review
Review Score
10/11
Quality Score
52
Canonical Locale
en
Detected Body Locale
en

适用场景: Ideal for AI agents that need erdos problem proving workflow. 本地化技能摘要: This workflow helps you prove Erdos problems in Lean 4 using your Claude Code/Codex subscription instead of paid API calls. It covers cli, combinatorics, erdos-problems workflows.

核心价值

推荐说明: erdos-prove helps agents erdos problem proving workflow. This workflow helps you prove Erdos problems in Lean 4 using your Claude Code/Codex subscription instead of paid API calls.

适用 Agent 类型

适用场景: Ideal for AI agents that need erdos problem proving workflow.

赋予的主要能力 · erdos-prove

适用任务: Applying Erdos Problem Proving Workflow
适用任务: Applying Target Problem: $ARGUMENTS
适用任务: Applying Tip: If no problem ID provided, check CANDIDATES.md for the current focus problem (see Decision

! 使用限制与门槛

  • 限制说明: Optional API sources (rate-limited, mostly free):
  • 限制说明: uv run erdos refs s2 citations <doi - Semantic Scholar (rate-limited)
  • 限制说明: Requires repository-specific context from the skill documentation

Why this page is reference-only

  • - Current locale does not satisfy the locale-governance contract.

Source Boundary

The section below is imported from the upstream repository and should be treated as secondary evidence. Use the Killer-Skills review above as the primary layer for fit, risk, and installation decisions.

评审后的下一步

先决定动作,再继续看上游仓库材料

Killer-Skills 的主价值不应该停在“帮你打开仓库说明”,而是先帮你判断这项技能是否值得安装、是否应该回到可信集合复核,以及是否已经进入工作流落地阶段。

实验室 Demo

Browser Sandbox Environment

⚡️ Ready to unleash?

Experience this Agent in a zero-setup browser environment powered by WebContainers. No installation required.

Boot Container Sandbox

常见问题与安装步骤

以下问题与步骤与页面结构化数据保持一致,便于搜索引擎理解页面内容。

? FAQ

erdos-prove 是什么?

适用场景: Ideal for AI agents that need erdos problem proving workflow. 本地化技能摘要: This workflow helps you prove Erdos problems in Lean 4 using your Claude Code/Codex subscription instead of paid API calls. It covers cli, combinatorics, erdos-problems workflows.

如何安装 erdos-prove?

运行命令:npx killer-skills add The-Obstacle-Is-The-Way/erdos-banger。支持 Cursor、Windsurf、VS Code、Claude Code 等 19+ IDE/Agent。

erdos-prove 适用于哪些场景?

典型场景包括:适用任务: Applying Erdos Problem Proving Workflow、适用任务: Applying Target Problem: $ARGUMENTS、适用任务: Applying Tip: If no problem ID provided, check CANDIDATES.md for the current focus problem (see Decision。

erdos-prove 支持哪些 IDE 或 Agent?

该技能兼容 Cursor, Windsurf, VS Code, Trae, Claude Code, OpenClaw, Aider, Codex, OpenCode, Goose, Cline, Roo Code, Kiro, Augment Code, Continue, GitHub Copilot, Sourcegraph Cody, and Amazon Q Developer。可使用 Killer-Skills CLI 一条命令通用安装。

erdos-prove 有哪些限制?

限制说明: Optional API sources (rate-limited, mostly free):;限制说明: uv run erdos refs s2 citations <doi - Semantic Scholar (rate-limited);限制说明: Requires repository-specific context from the skill documentation。

安装步骤

  1. 1. 打开终端

    在你的项目目录中打开终端或命令行。

  2. 2. 执行安装命令

    运行:npx killer-skills add The-Obstacle-Is-The-Way/erdos-banger。CLI 会自动识别 IDE 或 AI Agent 并完成配置。

  3. 3. 开始使用技能

    erdos-prove 已启用,可立即在当前项目中调用。

! 参考页模式

此页面仍可作为安装与查阅参考,但 Killer-Skills 不再把它视为主要可索引落地页。请优先阅读上方评审结论,再决定是否继续查看上游仓库说明。

Upstream Repository Material

The section below is imported from the upstream repository and should be treated as secondary evidence. Use the Killer-Skills review above as the primary layer for fit, risk, and installation decisions.

Upstream Source

erdos-prove

安装 erdos-prove,这是一款面向AI agent workflows and automation的 AI Agent Skill。查看评审结论、使用场景与安装路径。

SKILL.md
Readonly
Upstream Repository Material
The section below is imported from the upstream repository and should be treated as secondary evidence. Use the Killer-Skills review above as the primary layer for fit, risk, and installation decisions.
Supporting Evidence

Erdos Problem Proving Workflow

Target Problem: $ARGUMENTS

Tip: If no problem ID provided, check CANDIDATES.md for the current focus problem (see Decision Log).

This workflow helps you prove Erdos problems in Lean 4 using your Claude Code/Codex subscription instead of paid API calls.

Overview

text
1┌─────────────────────────────────────────────────────────────────┐ 2│ COST-FREE PROVING WORKFLOW │ 3├─────────────────────────────────────────────────────────────────┤ 4│ 1. Understand → Read problem statement (FREE) │ 5│ 2. Research → Gather literature context (FREE) │ 6│ 3. Formalize → Generate Lean skeleton (FREE) │ 7│ 4. Prove → Work with Claude Code (SUB) │ 8│ 5. Verify → Check Lean compilation (FREE) │ 9│ 6. Iterate → Fix errors with Claude Code (SUB) │ 10└─────────────────────────────────────────────────────────────────┘

Step 1: Understand the Problem

First, let me read the problem statement and understand what we're trying to prove.

Action: I will read the problem details using:

  • data/problems_enriched.yaml (if present) or src/erdos/data/problems_enriched.yaml (built-in sample dataset)
  • erdos show $ARGUMENTS output for formatted display

Questions to answer:

  • What is the mathematical statement?
  • What is the current status (open, partially solved)?
  • Are there any known partial results?

Step 2: Research Existing Literature

Gather context from existing references without making API calls.

Local sources (FREE):

  • literature/manifests/$(printf '%04d' $ARGUMENTS).yaml - Reference metadata (IDs are zero-padded, e.g. 6 → 0006.yaml)
  • literature/cache/ - Downloaded papers and sources
  • research/problems/$(printf '%04d' $ARGUMENTS)/ - Research workspace (leads, hypotheses, tasks)
  • Search index: uv run erdos search "relevant terms" --problem $ARGUMENTS (omit --problem to search globally)

Optional API sources (rate-limited, mostly free):

  • uv run erdos refs zbmath --msc <relevant-code> - zbMATH (free)
  • uv run erdos refs s2 citations <doi> - Semantic Scholar (rate-limited)
  • uv run erdos ingest $ARGUMENTS --source openalex - Fetch refs by DOI/arXiv ID (free)

Optional AI-powered discovery (PAID):

  • uv run erdos research exa search $ARGUMENTS "relevant query" --save-leads - Exa AI search

Step 3: Generate Lean Skeleton

Create the formal statement file:

bash
1uv run erdos lean formalize $ARGUMENTS

This creates: formal/lean/Erdos/Problem$(printf '%03d' $ARGUMENTS).lean (IDs are zero-padded, e.g. 6 → Problem006.lean)

What this generates:

  • Import statements for Mathlib
  • Formal theorem statement (with sorry placeholder)
  • Comments with problem context

Step 4: Develop the Proof (SUBSCRIPTION)

Now we work together to fill in the proof. This is where your subscription pays off instead of API calls.

I will directly edit these files (no copy/paste needed):

  • formal/lean/Erdos/Problem$(printf '%03d' $ARGUMENTS).lean - The main proof file
  • data/problems_enriched.yaml - Update status when solved

My workflow:

  1. Read the generated Lean file
  2. Analyze the theorem statement
  3. Propose proof strategies
  4. Write Lean tactics step by step
  5. Edit the file directly - you just watch and run lean check

Proof development strategies:

  • Break into lemmas if complex
  • Use Mathlib tactics (simp, ring, norm_num, omega)
  • Apply relevant theorems from Mathlib
  • Build incrementally, checking each step

Step 5: Verify Compilation

Check that the proof compiles:

bash
1PROBLEM3=$(printf '%03d' $ARGUMENTS) 2uv run erdos lean check "formal/lean/Erdos/Problem${PROBLEM3}.lean"

Possible outcomes:

  • Success: Proof compiles, no sorry remaining
  • Errors: Type mismatches, tactic failures, missing imports
  • Warnings: Unused variables, style issues

Step 6: Iterate on Errors (SUBSCRIPTION)

When compilation fails, show me the error output and I'll:

  1. Diagnose the issue
  2. Propose fixes
  3. Edit the file directly
  4. Repeat until it compiles

Common fixes:

  • Import missing Mathlib modules
  • Adjust tactic arguments
  • Add intermediate steps
  • Fix type annotations

Workflow Commands Reference

bash
1# Generate skeleton (if not done) 2uv run erdos lean formalize $ARGUMENTS 3 4# Check compilation 5PROBLEM3=$(printf '%03d' $ARGUMENTS) 6uv run erdos lean check "formal/lean/Erdos/Problem${PROBLEM3}.lean" 7 8# View problem details 9uv run erdos show $ARGUMENTS 10 11# Search related literature 12uv run erdos search "relevant terms" --problem $ARGUMENTS # omit --problem to search globally 13 14# Check formalization status 15uv run erdos lean status $ARGUMENTS 16 17# View execution logs 18uv run erdos logs --problem-id $ARGUMENTS

What I Can Do For You

Read & Analyze:

  • Problem statements from YAML
  • Lean files (existing proofs, Mathlib source)
  • Literature (PDFs, markdown)
  • Error messages

Edit & Write:

  • Lean proof tactics
  • Helper lemmas
  • Type annotations
  • Import statements

Research:

  • Find relevant Mathlib theorems
  • Explore proof strategies
  • Identify useful tactics

Example Conversation

text
1You: /erdos-prove 6 2 3Claude: Let me start the proving workflow for Problem 6. 4 5[Step 1] Reading problem statement... 6Problem 6 is about... [explains math] 7 8[Step 2] Checking existing literature... 9Found 3 references in the manifest: [lists them] 10 11[Step 3] I see formal/lean/Erdos/Problem006.lean exists. Let me read it... 12Current state: Has `sorry` at line 42. 13 14[Step 4] Here's my proposed proof approach: 151. We can use the pigeonhole principle 162. Apply Mathlib's Finset.card_le_card 173. Conclude with norm_num 18 19Let me edit the file to add the proof... 20[Makes edit] 21 22[Step 5] Now run: uv run erdos lean check formal/lean/Erdos/Problem006.lean 23 24You: [pastes error output] 25 26Claude: I see the issue - we need to import Data.Nat.Prime. 27Let me fix that... 28[Makes edit] 29 30Try again: uv run erdos lean check formal/lean/Erdos/Problem006.lean

Why This Saves Money

Traditional ApproachSubscription Approach
erdos loop run 6Work with Claude Code
~$0.50-5.00 per run$0 (subscription)
Automated but costlyInteractive but free
Limited iterationsUnlimited iterations

Alternative: Aristotle API (Paid)

If you prefer hands-off automated proving via Harmonic's Aristotle API:

bash
1# 1. Install aristotlelib 2uv sync --extra aristotle 3 4# 2. Set API key (use export for direct CLI calls) 5export ARISTOTLE_API_KEY=arstl-your-key 6 7# 3. Run via erdos wrapper (recommended - auto-loads .env) 8uv run erdos lean prove formal/lean/Erdos/Problem006.lean \ 9 --output formal/lean/Erdos/Problem006_aristotle.lean 10 11# Or direct CLI (requires exported env var) 12uv run aristotle prove-from-file \ 13 formal/lean/Erdos/Problem006.lean \ 14 --output-file formal/lean/Erdos/Problem006_aristotle.lean

Cost: Paid per-problem. Use the subscription workflow above for unlimited iterations.

Ready to Begin?

Tell me to start and I'll begin Step 1: reading and understanding Problem $ARGUMENTS.

Or if you want to jump to a specific step:

  • "Skip to step 3, the skeleton already exists"
  • "Start from step 4, I have the Lean file ready"
  • "I have errors, help me with step 6"

相关技能

寻找 erdos-prove 的替代方案 (Alternative) 或可搭配使用的同类 community Skill?探索以下相关开源技能。

查看全部

openclaw-release-maintainer

Logo of openclaw
openclaw

本地化技能摘要: 🦞 # OpenClaw Release Maintainer Use this skill for release and publish-time workflow. It covers ai, assistant, crustacean workflows. This AI agent skill supports Claude Code, Cursor, and Windsurf workflows.

333.8k
0
AI

widget-generator

Logo of f
f

本地化技能摘要: Generate customizable widget plugins for the prompts.chat feed system # Widget Generator Skill This skill guides creation of widget plugins for prompts.chat . It covers ai, artificial-intelligence, awesome-list workflows. This AI agent skill supports Claude Code, Cursor, and Windsurf

149.6k
0
AI

flags

Logo of vercel
vercel

本地化技能摘要: The React Framework # Feature Flags Use this skill when adding or changing framework feature flags in Next.js internals. It covers blog, browser, compiler workflows. This AI agent skill supports Claude Code, Cursor, and Windsurf workflows.

138.4k
0
浏览器

pr-review

Logo of pytorch
pytorch

本地化技能摘要: Usage Modes No Argument If the user invokes /pr-review with no arguments, do not perform a review . It covers autograd, deep-learning, gpu workflows. This AI agent skill supports Claude Code, Cursor, and Windsurf workflows.

98.6k
0
开发者工具