Remote Sensing Nature Skills

This repository collects reusable Codex/agent skills for remote-sensing academic workflows. The skills are focused on manuscript writing, polishing, paper-to-slides conversion, paper reading, reviewer-style assessment, literature search, and publication-grade scientific figures.

Repository Layout

skills/
  _shared/
  remote-sensing-nature-academic-search/
  remote-sensing-nature-figure/
  remote-sensing-nature-paper2ppt/
  remote-sensing-nature-polishing/
  remote-sensing-nature-reader/
  remote-sensing-nature-reviewer/

Each skill contains a SKILL.md router, a manifest.yaml when needed, and supporting static/, references/, scripts/, assets/, or evals/ folders.

skills/_shared/ contains shared writing, terminology, ethics, and journal-format fragments used by several skills. Keep it next to the skill folders so relative paths such as ../_shared/core/terminology-ledger.md continue to work.

Quick Install

Copy the skill folders into your local Codex skills directory:

mkdir -p ~/.codex/skills
cp -R skills/remote-sensing-nature-* ~/.codex/skills/

Then restart Codex or refresh your skill registry.

Skills

Typical Prompts

Use remote-sensing-nature-reader to read the first 8 pages of this PDF and create a Chinese-English Markdown reader.

Use remote-sensing-nature-paper2ppt to make a Chinese group-meeting PPT from this remote-sensing benchmark paper.

Use remote-sensing-nature-figure with Python to create a Nature-style multi-panel figure from this CSV.

Use remote-sensing-nature-polishing to polish this abstract for a remote-sensing AI paper.

Design Pattern

Most skills use a router structure:

1. Read SKILL.md.
2. Load manifest.yaml.
3. Load only the core fragments and workflow fragments needed for the current request.
4. Use deeper references/ files only when the task requires them.

This keeps each invocation focused while preserving detailed guidance for complex tasks.

Notes

• The skills are designed for remote sensing, Earth observation, geospatial AI, GIScience, photogrammetry, MLLMs, multimodal benchmarks, and related academic workflows.
• They emphasize traceability and non-invention: do not fabricate citations, results, repository IDs, reviewer comments, figure panels, or line numbers.
• Some workflows may require local tools such as Python, matplotlib, python-pptx, PyMuPDF, pypdf, or R packages depending on the selected backend.