nsfc-roadmap_skill

Overview

This skill generates publication-quality NSFC technical roadmaps and printable A4-ready flowcharts from proposal content or a structured spec. It produces an editable draw.io source (.drawio) plus embeddable outputs (.svg, .png, .pdf) and a reviewable planning markdown. Use it when you need a reproducible, review-friendly visual roadmap rather than minor image edits or text polishing.

How this skill works

The tool ingests either a proposal file/directory or a user-supplied spec and drafts a validated spec if needed. It renders the roadmap through a deterministic pipeline that produces draw.io source, vector and raster exports, and an iteration record with multi-round structural, visual, and readability evaluations. Optional AI-guided planning or AI critic loops are supported through a controlled request/response protocol.

When to use it

• You need a printable, A4-legible technical roadmap derived from proposal research content.
• You want an editable source (.drawio) plus embeddable assets (.svg/.png/.pdf) for reports or LaTeX/Word insertion.
• You prefer a reproducible spec-driven workflow for reviewers and collaborators.
• You want structured iterative improvements with documented critiques and a best-round artifact.
• You require the roadmap to follow explicit layout templates (classic/three-column/layered-pipeline).

Best practices

• Provide either proposal_file/proposal_path or a spec_file; a ready spec yields the most predictable output.
• Start with the planning stage (roadmap-plan.md) to lock template_ref and node density limits before rendering.
• Limit node density: aim for 3–5 phases, 2–6 nodes per phase to keep A4 readability.
• Keep terminology consistent between text and spec to avoid duplicate concepts in nodes.
• Use ai planning or ai_critic only when you will inspect and approve generated spec patches; follow the minimal response protocol for reproducibility.

Example use cases

• Convert 2.1.研究内容.tex sections into a three-column technical roadmap for a funding submission.
• Generate an editable draw.io roadmap and high-resolution PNG/PDF for inclusion in a grant PDF or presentation.
• Run 5 optimization rounds to iteratively improve layout, contrast, and node readability before final submission.
• Produce a reproducible output package (spec_latest.yaml, optimization_report.md) to share with co-authors for review.
• Enable ai_critic to solicit structured improvement suggestions and apply the best automated patch in the next round.