nsfc-code_skill

Overview

This skill recommends five NSFC application code pairs (primary/secondary) based on the text of an NSFC proposal and the 2026 recommendation library. It reads the proposal in read-only mode, ranks candidate codes by text similarity, and produces a timestamped Markdown report with rationale tied to both proposal excerpts and recommendation descriptions. The tool never modifies user files and only suggests codes present in the reference TOML.

How this skill works

It recursively reads the provided proposal directory or main .tex file (ignoring build artifacts) and extracts topic, methods, objects, and keywords as a read-only analysis. A deterministic ranking script scores each code's 'recommend' description against the extracted text to create a candidate list. From that list, the skill selects five pairs of codes (primary and secondary) and writes a Markdown file NSFC-CODE-vYYYYMMDDHHmm.md containing the five recommendations and traceable reasons that cite both proposal keywords and library descriptions.

When to use it

• When you have an NSFC proposal text but are unsure which application codes to pick.
• When you need 5 defensible code pairs with explicit textual justification.
• When you want recommendations strictly limited to the 2026 recommendation library.
• Before submitting to check alternative secondary codes for review strategy.
• When you prefer an audit trail linking proposal content to code descriptions.

Best practices

• Provide the main .tex path or the proposal directory so the skill can extract full context.
• Optionally state whether you prefer theoretical, methods, engineering, cross-disciplinary, or translational emphasis.
• If you know the broad division (e.g., class A), supply a prefix filter to reduce noise.
• Review highlighted proposal keywords in the report and confirm any ambiguous interpretations.
• Treat the output as recommendations to validate with domain colleagues or program officers.

Example use cases

• A junior PI preparing a young scientist application who needs candidate main and backup codes.
• A team rewriting methodology to better align with a target disciplinary code.
• A proposal editor checking consistency between research aims and chosen NSFC codes.
• A researcher comparing multiple candidate codes to plan submission strategy.
• An administrative assistant generating a code recommendation file for PI review.