I open-sourced a set of MathModeling-skills!

zhnnky.

Learn how to install and use MathModeling-skills to run a 24-step math modeling workflow. This includes using install.sh to install native Claude Code and Codex plugins or enabling project-level Skills, following steps from problem parsing → literature analysis → method generation → Python/MATLAB code generation → code review → experiment reporting → robustness analysis → figure planning → paper writing and polishing. See README and CLAUDE.md for installation and configuration details (confirm repo access).

Image Analysis

Image 1 — Recorded facts (only visible evidence summarized)

1. Overall interface and visible elements (facts)

- Platform view: GitHub repository page, dark mode.

- Top navigation (left to right readable text/icons):

- Repository path: KyrieZhang329 / MathModeling-skills (repo name shows a lock icon to the right).

- Top menu items: Code, Issues, Pull requests, Agents, Actions, Projects, Wiki, Security and quality, Insights, Settings (all visible tabs).

- Top-right search box placeholder text: Type / to search.

- Several small icons at top-right (not all readable as text; only graphical buttons and avatar visible).

- Repo header area:

- Large title: MathModeling-skills (with a gray “Private” badge at right).

- At right, three statistic buttons: Watch 0, Fork 0, Star 0.

- Branch/tag area:

- Branch selector shows: main.

- Text: 1 Branch, 0 Tags.

- Center has a "Go to file" search box (placeholder readable).

- Right-side buttons: Add file (gray), and a green <> Code button.

- Commit row above file list:

- Collaborator/commit line: KyrieZhang329 and claude synced all docs to 24 skill latest workflow (right shows commit id a1cc816 • 3 minutes ago, and 18 Commits).

- File/directory list (in order, left column file names, right column latest commit info and times, all visible facts):

1. .claude/skills — added 4 skills + updated code generator + sync .codex/skills — 18 minutes ago

2. .codex/skills — added 4 skills + updated code generator + sync .codex/skills — 18 minutes ago

3. docs — synced all docs to 24 skill latest workflow — 3 minutes ago

4. examples — added repo framework and template files — 3 weeks ago

5. skills-docs — synced all docs to 24 skill latest workflow — 3 minutes ago

6. templates — refactored modeling skills workflow, formally added paper analysis stage, will model-c... (commit message truncated) — 2 weeks ago

7. .gitignore — initial first-version skills docs completed — 3 weeks ago

8. AGENTS.md — fully refactored skills workflow per LYD problem analysis and AI modification suggestions — 50 minutes ago

9. CLAUDE.md — updated to reflect full workflow and 24 skills — 10 minutes ago

10. CONTRIBUTING.md — added repo framework and template files — 3 weeks ago

11. Initial Prompt-zh.md — synced all docs to 24 skill latest workflow — 3 minutes ago

12. Initial Prompt.md — synced all docs to 24 skill latest workflow — 3 minutes ago

13. LICENSE — Initial commit — 3 weeks ago (file name and commit message visible)

(List continues downward; more files are cut off by screenshot)

- Right-side About panel (readable facts):

- About description (in Chinese): Claude Code / Codex Skills for math modeling competitions, supporting phased modeling workflow and Python, MATLAB/Beida Tanyuan code branches.

- Below: links/items: Readme, MIT license, Contributing, Activity.

- Repo stats visible: 0 stars, 0 watching, 0 forks.

- Releases: No releases published; link text Create a new release.

- Packages: No packages published; link text Publish your first package.

- Contributors line partially visible (text “Contributors 1” visible near bottom).

- Other visible observations:

- Typical GitHub repo main view (file list + right sidebar).

- There is a scrollbar; page scrolled to show some files; a translucent circular button at bottom-right (unclear, possibly a floating control).

2. Unreadable or truncated content (facts)

- The commit message for templates directory is truncated, showing “will model-c...” with remainder not visible.

- Several small top-right icons lack textual labels and cannot be fully identified.

- More files appear below in file list but are cropped, preventing reading their names and commit messages.

- Some lines at bottom of About panel and page footer are partially cropped and not fully readable.

3. Inferences (marked as inferences based on visible info)

- Inference 1: The repository appears set to private (shows Private badge). This is inferred from UI but not an absolute backend confirmation.

- Inference 2: The repo appears to collaborate with a user or bot named “claude” (UI shows “KyrieZhang329 and claude”), suggesting automated synchronization.

- Inference 3: The repo aims to provide staged modeling skills templates and code examples for math modeling competitions (based on About text and file/commit names).

- Inference 4: Presence of Initial Prompt-zh.md and Initial Prompt.md suggests both Chinese and English initial prompt documents are maintained.

- Inference 5: Commit timestamps like “3 minutes ago / 18 minutes ago / 3 weeks ago” indicate frequent updates around the screenshot time.

4. Ignored content

If needed, I can export the file names with their commit messages as a text list, or perform a more detailed verbatim transcription for a specific line (truncated parts will be annotated as not visible).

SkillPI Creation Analysis

1) Obtain repository access (image shows repo may be private — request access if necessary).

2) Read README, CLAUDE.md and docs/skills-docs to learn the definitions and I/O for the 24 skills.

3) For native-plugin installation, run install.sh as provided (source states install.sh completes installation for native plugins); alternatively follow the project-level Skills deployment instructions in the repo.

4) Prepare runtime environments for Python and/or MATLAB per repository examples.

5) Invoke skills in sequence (problem parsing → literature analysis → method pool → code generation → review → reporting etc.) using the repo’s templates and prompts.

6) Apply the repository’s figure-planning approach (inspired by nature-skills’ figure contract) and the 12-step polishing checklist for manuscript polishing. Reproduction prerequisites: access to the repository (may be private), access to Claude Code or Codex plugin environments, a compatible OS and shell to run install.sh, and local/remote Python and/or MATLAB runtime. Items to verify before attempting reproduction:

- Repository access level (private vs public) and any invitation/permission requirements.

- install.sh requirements and supported Claude Code / Codex versions, OS/permission needs.

- Exact skill input/output formats and invocation API (details should be in docs or skills-docs).

- Details of the MATLAB “Beida Tanyuan” code branch and its compatibility/licensing.

- Any required API keys, paid services, or external dependencies not visible in the screenshot. Limitations: This interpretation is based on the provided summary and screenshot text; the install script, plugins, and actual runtime behavior were not executed or validated here. Verify details in the repository docs and test in a controlled environment.

Original creation ↗