What problem does it solve?
Writing a Chinese NSFC (National Natural Science Foundation) grant proposal requires meeting strict length constraints—especially the 2026+ rule limiting the main text to roughly 30 pages. Manually judging whether sections like the research basis, research content, or preliminary work are too long or too short is subjective and error-prone. This Skill turns length into a measurable, closed-loop metric: it checks your proposal against configurable budgets, reports deviations per file and per section, and guides meaning-preserving expansion or compression until the draft complies.
Core Features & Use Cases
- Deterministic length checking: A Python script counts CJK characters in LaTeX sources, optionally follows the
main.tex \input/\include dependency tree (ignoring commented-out inputs), and can count PDF pages as a hard constraint.
- Configurable budgets:
config.yaml defines per-file budgets (e.g., 立项依据 8000–10000 chars, 研究内容 12000–15000 chars) plus overall and page limits, aligned with 2026 NSFC guidelines.
- Closed-loop rewriting guidance: Structured Markdown/JSON reports with file-level and section-level deltas drive targeted expansion or compression, followed by mandatory re-checking.
- Use Case: You have a proposal built on the NSFC_Young template. Point the checker at the project root, receive a report showing the research content section is 3000 characters over budget, compress the redundant passages, then re-run the check to confirm compliance.
Quick Start
Ask the AI to run the length check on your NSFC proposal directory using the skill's config.yaml, read the generated length report, and suggest which sections to expand or compress to meet the budget.
