Skip to content
Agentmart
Submit a tool

wayfinder-mathmodel

Use when a mathematical modeling problem must be solved from its statement and attachments into a high-quality Chinese CUMCM paper: plan and execute through a…

Skill by Antrairs

IndexedApache-2.0
Write a review

Chalk Labs

Need help deploying wayfinder-mathmodel?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code. Read the repository before you give any tool access to your data.

npx agentmarket install skill/antrairs-wayfinder-mathmodel

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use when a mathematical modeling problem must be solved from its statement and attachments into a high-quality Chinese CUMCM paper: plan and execute through a Wayfinder map, keep Markdown as the source of truth, run real Python models, and deliver a verified LaTeX PDF.

Use cases: documents, research

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Python package

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/wayfinder-mathmodel

Tools it pre-approves 9

From allowed-tools in SKILL.md: the agent may use these without asking you each time.

  • Bash(*),
  • Read,
  • Write,
  • Edit,
  • Grep,
  • Glob,
  • Agent,
  • WebSearch,
  • WebFetch

What's inside

Skill name
wayfinder-mathmodel
Description
Use when a mathematical modeling problem must be solved from its statement and attachments into a high-quality Chinese CUMCM paper: plan and execute through a Wayfinder map, keep Markdown as the source of truth, run real Python models, and deliver a verified LaTeX PDF.
License
Apache-2.0
Allowed tools
Bash(*),, Read,, Write,, Edit,, Grep,, Glob,, Agent,, WebSearch,, WebFetch
Instructions
141 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts
  • Pre-approves some shell commands

Files (15)

  • SKILL.md
  • openai.yaml · agents/
  • external-resolver.md · references/
  • paper-structure.md · references/
  • result-contract.md · references/
  • wayfinder-contract.md · references/
  • check_claims.py · scripts/
  • check_pdf.sh · scripts/
  • compile_pdf.sh · scripts/
  • render_latex.py · scripts/
  • data.csv · scripts/tests/fixtures/cumcm_2024_c_smoke/
  • problem.md · scripts/tests/fixtures/cumcm_2024_c_smoke/
  • solve.py · scripts/tests/fixtures/cumcm_2024_c_smoke/
  • test_wayfinder.py · scripts/tests/
  • wayfinder.py · scripts/

SKILL.md (first lines)

# Wayfinder 数学建模

这是 MathModelAgent 的唯一用户入口。它把题面理解、建模、计算、写作和验收组织成一个可恢复的 Wayfinder 决策图;用户不需要依次调用内部 skill。规划和执行属于同一次任务:首次调用创建地图,之后每次调用从现有地图继续,直到交付 PDF、遇到必须由用户决定的问题,或明确阻塞。

## Destination

默认目的地是一个可提交的中文 CUMCM 数学建模论文工作区:

- paper/main.pdf:最终 PDF;
- paper/main.tex 与 paper/sections/*.tex:唯一排版源;
- paper/structure.md 与 results/*.md:论文结构和事实源;
- code/、figures/、problem_files/:可复核的输入、程序和证据。

正文必须是正式论文的一部分:不出现 AI、agent、ticket、草稿、占位符、内部流程或“不能作为因果证据”一类元话语。证据不足时,改写成证据支持的统计/预测/关联结论,不把免责声明塞进正文。

## First call and resume

1. 读取用户给出的题目正文和附件;Codex 能直接读取的文件直接使用,不强制引入 OCR 或文档摄取平台。
2. 以题目目录为 ROOT,运行 scripts/wayfinder.py init ROOT;已有 map.md 时只恢复,不覆盖。
3. map.md 只保留 Wayfinder 的五个地图区段:Destination、Notes、Decisions so far、Not yet specified、Out of scope。地图是索引;详细问题、答案和资产链接在 issues/。
4. 从题面生成首批子问题和必要决策票据,计算 frontier,认领一个可执行票据并处理它。研究票据可以并行;其他票据遵守“一次会话一个票据”。
5. 票据解决后立刻写回答案/资产链接、更新地图和 frontier;重新进入本 skill 时从未解决的已认领票据或 frontier 继续。

工作区固定为:

    ROOT/
    ├── map.md
    ├── issues/
    ├── problem_files/
    ├── code/
    ├── results/
    ├── figures/
    └── paper/
        ├── structure.md
        ├── sections/
        ├── main.tex
        └── main.pdf

## Wayfinder ticket contract

只使用四种票据类型:research、prototype、grilling、task。不要发明 decision、validation 或自定义状态。

- 票据正文主要只有 Question 段;答案写在解决时的 Resolution 段,资产用相对链接,不把大段数据粘进票据。
- claim 先写 assignee,再开始工作。depends_on 决定阻塞关系;未关闭依赖的票据不在 frontier。
- research 是 AFK 候选,可派生一个短生命周期研究上下文,按 Wayfinder 约定使用 research/name throwaway branch/context pointer;子上下文只返回研究结果和引用,父票据负责归并和解决。
- prototype 与 grilling 是 HITL 候选;task 可根据风险决定 AFK 或 HITL。
- 关闭票据只意味着该问题的答案或证据已经写入;重要解释仍需在论文源中用正式文字表达。

用脚本保持这些语义:

    python scripts/wayfinder.py init ROOT --destination "..."
    python scripts/wayfinder.py new ROOT --title "..." --type research --question "..."
    python scripts/wayfinder.py frontier ROOT
    python scripts/wayfinder.py claim ROOT --title "..." --assignee codex
    python scripts/wayfinder.py resolve ROOT --title "..." --resolution-file answer.md

面向用户显示票据标题,不显示裸 ID。失败的建模路线不删除:另建一个有依赖关系的新票据,在原票据答案中记录失败原因。

## Input gate

题面、数据列名、单位、图表或附件若不可读或有会改变模型的歧义,立即在聊天中停止并指出文件名、页码/位置、具体问题和用户需要采取的动作。此时不建票据、不写猜测数据、不生成部分论文。用户补充后,从地图恢复。

对能读出的事实由 agent 处理;只把不可约的人类选择交给用户,例如题意解释、关键假设、模型路线、最终发布。每次 HITL 问题都遵守零上下文原则:短、独立、面向初学者,并附推荐项:

    ❓ Q - 采用哪种题意解释?
    题面中的“覆盖”可能指 A(至少一次)或 B(累计次数)。
    A:……;B:……。
    ➡️ 建议:A,因为它与题目给出的评价指标一致。

用户回复 y 表示接受当前所有清晰的推荐;有歧义或高副作用时只追问一个聚焦问题。不要要求用户复述前文。

## Modeling and computation

把每个有实质影响的未知点拆成票据,而不是预先强行套固定六阶段。通常按以下条件调用内部解析器:

1. 2analysis-modeling:将当前问题闭合为变量、假设、目标、约束、求解方法和校验口径;最多保留两个候选模型族,先按正确性、可解释性、可计算性、验证性和稳定性比较,再考虑复杂度/新颖性。
2. 3coding-visual:只实现已经认领的建模票据,使用 Python 和仓库已有依赖;程序必须实际运行,优化先检验可行性,预测有验证划分或误差评价,图表必须由数据生成。
3. 4drawio:仅在整体思路或机制图能帮助读者理解时调用;不为装饰生成图。
4. 5writing:从 paper/structure.md 和结果事实生成 paper/sections/*.tex,只使用 LaTeX/CUMCM 模板;main.tex 是生成物,不手工维护。
5. 6verity:执行文本、引用、数值、代码和 LaTeX/PDF 检查;失败则回到产生该证据的票据,不用验收阶段掩盖问题。

默认使用“最简单的充分模型”。只有题目要求或验证显示收益明确时才升级复杂度。研究/原型的目标是消除不确定性,不是增加方法数量。
若宿主提供外部 GitHub 数学建模 skill,遵循 references/external-resolver.md:当前票据最多绑定一个最匹配的外部 resolver,完成后仍由本 skill 归并和验收。

## Evidence contract

results/ 是数值和结论的事实源。每条论文级结论至少有一个 Markdown 结果条目,字段见 references/result-contract.md:

    ---
    claim: "..."
    value: "..."
    unit: "..."
    method: "..."
    source_artifact: "code/solve.py"
    validation: "..."
    execution_status: "passed"
    run_command: "python code/solve.py"
    reported_value: "..."
    ---

禁止编造、手算后伪装成运行结果或用未记录的四舍五入替代原值。随机算法记录 seed;运行命令在不明显时记录。图表、表格和正文数字都必须能追溯到结果条目及其源资产。

## Adaptive paper contract

先读 references/paper-structure.md,再为当前题目生成 paper/structure.md。它是自适应章节图,不是固定目录。优秀 CUMCM 论文常见的可选骨架是:

- 摘要(问题、方法、关键数字、结论);
- 问题重述、问题分析、符号与假设;
- 数据处理/探索性分析(仅在有用时集中写);
- 每个子问题一个闭环:分析 → 假设/变量 → 模型 → 求解 → 结果 → 验证/解释;
- 模型评价、敏感性/稳健性、推广与局限;
- 参考文献、代码/数据附录。

不机械复制示例论文,也不因为题目没有三问而强行补三章。每张图表都有问题、数据源、生成程序、正文解释和 caption;没有证据支撑的图表不进入正文。论文正文先在 sections/*.tex 完成,模板只负责排版。

## Finish gates

只在以下条件同时满足时向用户交付“最终”:

From Antrairs/mathmodel-skills, Apache-2.0.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Aug 23, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from Antrairs

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get wayfinder-mathmodel running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying wayfinder-mathmodel?

CallGet help
Cite this page

wayfinder-mathmodel skill. Agent Mart. https://agentmart.dev/skills/antrairs-wayfinder-mathmodel (last updated 2026-10-24)