PaperForge v2.0.0 — runnable Agent Skill (Topic in. Paper out.)

🎉 PaperForge v2.0.0

Topic in. Paper out. — a natural-language research topic becomes an 8000+ word remote sensing manuscript (markdown + HTML + DOCX) with GEE-computed figures and tables.

What's new vs v1

v1	v2.0
Instructions only — agent re-derives every code block	✅ Reusable Python modules (gee_tasks.py, paper_writer.py, render_figures.py, literature.py)
No demo	✅ python scripts/demo.py → real figures + .md/.html/.docx in seconds — no GEE, no internet
No tests	✅ python scripts/test_pipeline.py — 9 offline unit checks
Manual Codex setup	✅ python scripts/package_codex.py --target ~/.codex/skills
Non-standard frontmatter	✅ Spec-compliant SKILL.md (name/description/license/compatibility/metadata/allowed-tools)
No release channel	✅ GitHub Releases for version pinning

🚀 Quick start

git clone https://github.com/xingguangYan/PaperForge.git
cd PaperForge
pip install -r requirements.txt
python scripts/demo.py          # offline smoke demo → outputs/demo/manuscript.{md,html,docx}

# Full pipeline with Google Earth Engine:
python scripts/run_pipeline.py --topic "Vegetation change in Wuhan" --project ee-you

✨ Capabilities

• 3-phase pipeline: ResearchX literature mining → GEEPro GEE execution → ResearchX manuscript generation
• Multi-round literature search, 15+ paper extraction, research-gap → GEE task design
• Automatic GeoTIFF download (all years/sites), statistics, 300-DPI publication figures
• 8000+ word manuscript with formulas, tables, 25-30 references (GB/T 7714 / APA / MLA)
• Cross-platform: Claude Code, OpenAI Codex CLI, Cline, Cursor, Windsurf, GitHub Copilot

✅ Verified

• test_pipeline.py → 9/9 pass
• demo.py → md + html + valid docx + figures
• run_pipeline.py --dry-run → full manuscript package

Full changelog: 93e73c1...v2.0.0