nsfc-code

Recommends five NSFC application code pairs from proposal text using heuristic similarity ranking.

2.7k|251|Updated Feb 12, 2024
One-click install
npx skills add https://github.com/huangwb8/ChineseResearchLaTeX --skill nsfc-code
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: nsfc-code
Source: https://github.com/huangwb8/ChineseResearchLaTeX/tree/main/skills/nsfc-code
Command: npx skills add https://github.com/huangwb8/ChineseResearchLaTeX --skill nsfc-code

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Choosing the correct NSFC (National Natural Science Foundation of China) application codes is difficult because the code library is large and interdisciplinary boundaries are blurry. This Skill reads your proposal text and produces five justified primary/secondary code recommendations without modifying any proposal files.

Core Features & Use Cases

  • Read-only proposal analysis: Recursively reads .tex/.md/.txt proposal content, strips LaTeX markup, and never modifies .tex/.bib/.cls/.sty files.
  • Deterministic candidate ranking: A dependency-free Python script scores proposal text against the nsfc_code_recommend.toml reference library using n-gram Jaccard similarity, with optional discipline-prefix filtering (e.g., --prefix A).
  • Traceable recommendations: Outputs five code1/code2 pairs with reasons citing both proposal keywords and the reference library's recommend descriptions, saved as NSFC-CODE-vYYYYMMDDHHmm.md.
  • Use Case: You have a LaTeX NSFC proposal on uncertainty quantification and robust control. Run the ranking script to shortlist 50 candidate codes, then let the AI finalize five justified primary/secondary code pairs in a timestamped Markdown report.

Quick Start

Analyze my NSFC proposal in projects/NSFC_Young and recommend five application code pairs with reasons, writing the result to a timestamped NSFC-CODE report.

Frequently Asked Questions about nsfc-code

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I choose NSFC application codes for my proposal?

Run the nsfc_code_rank.py script on your proposal directory to get a similarity-ranked candidate list, then have the AI select five primary/secondary code pairs with reasons. The final recommendations are written to a timestamped NSFC-CODE-vYYYYMMDDHHmm.md file.

What input formats does NSFC code recommendation support?

The skill accepts directories or single files containing .tex, .md, or .txt proposal content. LaTeX files are automatically stripped of comments, math, citations, and commands before similarity scoring against the code library.

Does the skill modify my LaTeX proposal files?

No, the skill is strictly read-only on proposal content. Guardrails forbid writing to any .tex, .bib, .cls, or .sty files; only the final Markdown report and isolated workspace intermediates are created.

How do I reduce cross-discipline noise in code ranking?

Use the --prefix flag to restrict candidates to a specific discipline letter, for example --prefix A for mathematics and physical sciences. This filters the reference library before similarity scoring, improving candidate relevance.

What are the limitations of automated NSFC code recommendation?

The heuristic n-gram ranking only narrows candidates; final selection requires AI semantic judgment and user confirmation. The skill is not optimized for the Interdisciplinary Science Department (T codes), so those results need expert review against official NSFC guidelines.