Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
PaperOrchestra provides a structured framework for AI research scientists to automate the drafting of academic papers. It transforms raw experimentation logs into formal manuscripts by iteratively refining LaTeX drafts through a simulated peer review process.
The system employs deterministic decision bands and a devil's advocate guard to ensure critical findings are resolved before a draft is accepted, maintaining detailed worklogs and snapshots for every iteration.
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Follow the documented setup, then try a first task.
Setup
Automated setup for supported agents
Installs: Complete nine-skill PaperOrchestra bundle, installed into repository-local and global agent skill directories. · Claude Code, Cursor, Windsurf, Antigravity, Gemini, Codex, OpenClaw, OpenCode, Factory Droid
Before you start
Git
Bash
Python 3.10+ with pip
pdflatex for pipeline PDF compilation (texlive-latex-recommended and texlive-latex-extra)
matplotlib
numpy
Run from: ~/paper-orchestra
In a terminal, replace the literal <this repo> placeholder with the PaperOrchestra repository URL, clone the repository into ~/paper-orchestra, enter that directory, and install the deterministic helper requirements.
git clone <this repo> ~/paper-orchestra
cd ~/paper-orchestra
pip install -r requirements.txt # deterministic helpers only
From the repository root, run the setup script in a terminal. It prompts for optional Semantic Scholar, Exa, and Tavily API keys and a PaperBanana executable path; leave a prompt blank to skip it. The script installs all nine skills into the documented local and global agent directories.
Installs: Complete nine-skill PaperOrchestra bundle, symlinked globally into ~/.claude/skills. · Claude Code
Before you start
Git
Python 3.10+ with pip
pdflatex for pipeline PDF compilation (texlive-latex-recommended and texlive-latex-extra)
matplotlib
numpy
A shell providing mkdir and ln
Claude Code
Run from: ~/paper-orchestra
In a terminal, replace the literal <this repo> placeholder with the PaperOrchestra repository URL, clone the repository into ~/paper-orchestra, enter that directory, and install the deterministic helper requirements.
git clone <this repo> ~/paper-orchestra
cd ~/paper-orchestra
pip install -r requirements.txt # deterministic helpers only
In a terminal, create Claude Code's global skill directory and symlink every documented PaperOrchestra skill into it.
mkdir -p ~/.claude/skills
for s in paper-orchestra outline-agent plotting-agent literature-review-agent \
section-writing-agent content-refinement-agent paper-writing-bench \
paper-autoraters agent-research-aggregator; do
ln -sf ~/paper-orchestra/skills/$s ~/.claude/skills/$s
done
Installs: Complete nine-skill PaperOrchestra bundle, symlinked into ~/.all-skills for a host already configured to discover that directory.
Before you start
Git
Python 3.10+ with pip
pdflatex for pipeline PDF compilation (texlive-latex-recommended and texlive-latex-extra)
matplotlib
numpy
A shell providing mkdir and ln
A host configured to discover ~/.all-skills
Run from: ~/paper-orchestra
In a terminal, replace the literal <this repo> placeholder with the PaperOrchestra repository URL, clone the repository into ~/paper-orchestra, enter that directory, and install the deterministic helper requirements.
git clone <this repo> ~/paper-orchestra
cd ~/paper-orchestra
pip install -r requirements.txt # deterministic helpers only
In a terminal, create ~/.all-skills and symlink every documented PaperOrchestra skill into it.
mkdir -p ~/.all-skills
for s in paper-orchestra outline-agent plotting-agent literature-review-agent \
section-writing-agent content-refinement-agent paper-writing-bench \
paper-autoraters agent-research-aggregator; do
ln -sf ~/paper-orchestra/skills/$s ~/.all-skills/$s
done
Use the outline-agent skill on [workspace path]. Read workspace/inputs/idea.md, experimental_log.md, template.tex, and conference_guidelines.md; generate workspace/outline.json with its plotting, literature-review, and section plans; validate the JSON; and stop before Steps 2 and 3.
Use the paper-writing-bench skill on [paper PDF or extracted markdown] for benchmark ID [paper_id]. As a first result, create bench/[paper_id]/idea_sparse.md with the four documented sections, stopping before empirical verification and excluding citations, URLs, author identities, LaTeX math, and experimental results.
Use the paper-autoraters skill's Literature Review Quality autorater on [paper PDF or text] with [reference average citation count]. Return the documented JSON assessment with axis_scores, penalties, summary, and overall_score; do not compare it with another paper.
Claude
Codex
Cursor
Gemini CLI
OpenClaw
Windsurf
OpenCode
Antigravity
Factory Droid
Cline