RS-PaperClaw skill
A pipeline that picks remote sensing papers from arXiv daily, reviews each in a GitHub Issue and publishes a daily digest.
- 253 ★ on the repository
- Last commit
Extended access
Why this level
- The GitHub token creates Issues in a repository, the model key spends money, and both secrets sit in `.env`.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/rs-paperclawDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The GitHub token creates Issues in a repository, the model key spends money, and both secrets sit in `.env`.
This is third-party code. Review the repository files before installing.
What it does
RS-PaperClaw fetches arXiv candidates each day, filters them by keywords and then with a model, creates a review of each paper in its own GitHub Issue (details, summary, Chinese abstract, tags, first-page previews, ten questions) and builds a daily digest, also as an Issue. The digest is copied to `daily_reports/YYYYMM/YYYYMMDD.md` and a short summary goes to DingTalk or Feishu. The agent entry is `skills/rs-paper-pipeline/`: `SKILL.md`, the `scripts/cli.py` command (`run`, `doctor`, `filter --dry-run`, `reconcile`) and a guide for retargeting the topic. GitHub Actions workflows set the schedule. The default model is `deepseek-v4-flash`, and any OpenAI Chat Completions compatible one works.
Who it is for. Remote sensing researchers and groups who want to run a paper feed like this in their own repository for their own topic.
Good fit when
- A summary of new arXiv work on a topic is wanted each morning without scanning the feed by hand.
- A paper is discussed in comments on its Issue and the notes stay in the history.
- The pipeline is retargeted to another topic by editing keywords and the filter prompt.
- Before a live run, selection for a chosen date is checked with `filter --dry-run`.
- The daily digest is delivered to colleagues in DingTalk or Feishu.
Not a fit when
- You need journals such as RSE or ISPRS: the README lists them as planned.
- The repository must not collect many Issues.
- There is no model key or GitHub token.
Example request
Run `python3 scripts/cli.py doctor`, then `run --date 20260317 --no-notify` and show the created Issues.Limitations
Two secrets are required, `GITHUB_TOKEN` and `LLM_API_KEY`: the token creates and updates Issues, the key pays for model calls. The only source for now is arXiv, with journals planned. Documentation and reviews are in Chinese. A full clone weighs over 400 MB because of previews, and `--filter=blob:none --sparse` gives a light copy. Python 3.10 or newer and `bootstrap.sh` are needed to run. Secrets live in a local `.env`, and the GitHub token must be allowed to write to the repository that receives the Issues.
How to disable. Disable the workflows in `.github/workflows/` and delete `GITHUB_TOKEN` and `LLM_API_KEY` from secrets. Created Issues and the `daily_reports/` directory stay and are removed by hand.
FAQ
Which secrets are needed?
`GITHUB_TOKEN` and `LLM_API_KEY` are required. The rest is optional: `LLM_MODEL`, `LLM_API_URL`, `DINGTALK_WEBHOOK`, `FEISHU_TARGET`, `RS_GITHUB_REPO`.
Can I change the model?
Yes. `LLM_MODEL` and `LLM_API_URL` accept any OpenAI Chat Completions compatible interface. The default is `deepseek-v4-flash`.
How do I change the topic?
Edit `scripts/config/filter_keywords.json` and `scripts/prompts/filter_cross_prompt.md`. `CUSTOMIZATION_GUIDE.md` covers the other steps and safe verification.
Where do results stay?
Issues hold the discussion and Markdown in `daily_reports/` serves as a static archive.
How often does it run?
Once a day: the main mode is a schedule in `.github/workflows/`, and a manual run has its own file there.
Discussion
No questions or reviews yet.
Related
SkillsA skill that finds AI papers from 28 labs on arXiv over any date range and splits led papers from co-authored ones.
Codex skill that picks fresh research papers for a researcher profile and builds a scored summary in Chinese.
An agent skill that turns an arxiv paper into a minimal Python implementation, with code anchored to sections and equations and gaps flagged honestly
An archive generator for Reddit, Voat and Ruqqus dumps on PostgreSQL with search and an MCP server of 29 tools for an agent
26 skills and four reviewer agents for science in Claude Code: figure and statistics review, literature search, work handoff.
A template for data scientists: AGENTS.md, CLAUDE.md and skills for Python, SQL, Polars and model diagnostics.
README in short
Storage is dual: Issues for discussion and Markdown for the archive. Selection runs in two steps, by keywords and then by model.