Data analysis skills specifically designed for the financial risk control field

[REAL-Madrid01]

Pull Request Checklist

• [√] I have read and followed the CONTRIBUTING.md guidelines.
• [√] My contribution adds a new skill file in the correct directory.
• [√] The file follows the required naming convention.
• [√] The content is clearly structured and follows the example format.
• [√] I have tested my instructions, prompt, agent, skill, or workflow with GitHub Copilot.
• [√] I have run npm start and verified that README.md is up to date.

---

Description

This is a skill that has been proven reliable in real scenarios. Its prototype comes from the code I wrote manually in my daily work. Now with the help of glm5, I have organized it into a skill, which is mainly used for credit risk data cleaning and variable screening pipelines for pre-loan modeling in financial scenarios. Use when working with raw credit data that requires quality assessment, missing value analysis, or variable selection prior to modeling. Covers data loading and formatting, abnormal cycle filtering, missing rate calculation, high missing variable removal, low IV variable filtering, high PSI variable removal, Null Importance denoising, high correlation variable removal, and cleaning reports.

---

Type of Contribution

• [√] New skill file.

---

Additional Notes

I have a dataset for rocks and a report generated accordingly. In order to match the specification of the skill and reduce invalid input tokens! it did not add here, anyone interested send me the msg in github i will be happy to offer that! furthermore, I am completing four plans in the future and hope to complete my project.

1. Add xgb and lr parameter adjustment. Currently, only lgb is optimized. The hyperopt_lgb.py file supports three parameter optimization methods of lgb xgb lr.
2. Support optuna parameter optimization. Only hyperopt optimization is obsolete. Use independent .py files to complete the support for xgb lgb lr parameter optimization function.
3. Support ensemble methods. Currently, it is a single model. Use existing ensemble methods and ensure over-fitting.
4. Supports table learning. Currently, only traditional methods are available. Add XFRM and tableNet.

---

By submitting this pull request, I confirm that my contribution abides by the Code of Conduct and will be licensed under the MIT License.

[aaronpowell]

it looks like you've branched off the main branch, not staged meaning you have all the materialised plugin files.

You can fix this with:

git fetch origin staged
git rebase --onto origin/staged origin/main <your branch name>
git push --force-with-lease

Or by using the script npm run plugin:clean

[aaronpowell]

We have a policy that items within this repo are English-first (we have done a few translations of existing instructions but they are translations rather than non-English exclusive).

The reason for this is that models are most optimised for English language parsing, and thus using the least amount of tokens for what needs to be done.

[aaronpowell]

Merging the main branch shouldn't be done as it brings the plugins folder across incorrectly. You're best to run npm run plugin:clean and probably need to run npm run build to get to the correct branch state.