Our take
Newless than six months old and changing fast- Activity
- 54
- Community
- 38
- Issues
- 53
- Pull requests
- 67
+16 stars in the last 30 days
AREX-Skill is a promising, actively developed skill library for agent-assisted research, but its young age and forthcoming full benchmark report call for evaluation on your own workflows.
Good fit if
- Your coding agent needs reusable procedures for ML repositories or research tasks.
- You want to route requests to focused skill instructions and export selected skills to Codex, Claude Code, or another compatible agent.
- You can work with DisCo and Node.js 22.19.0 or newer, and check the license metadata for each skill you use.
Look elsewhere if
- You need a mature, long-established platform rather than a project that is less than six months old and changing quickly.
- You need a complete technical report before relying on the published benchmark results.
- You need a single license to govern every included skill, since individual skills can have their own license terms.
All health signals
| Last commit | 2026-09-03 (1 month ago) |
|---|---|
| Commits, last 90 days | 60 |
| Releases, last 12 months | 1 (latest v0.2.1, 2026-09-02) |
| Contributors | 6 (top contributor: 70% of commits) |
| Issues closed, last 90 days | 2 |
| Pull requests merged, last 90 days | 18 (typically merged in 0 days) |
| Project age | 5 months |
Checked on 2026-10-16 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
354Forks
28Change since Nov 17, 2026: 0 stars·0 forks
Overview
AREX-Skill turns operating knowledge from GitHub repositories and papers into executable skills for coding agents and autonomous machine learning research. Built and used through DisCo, its CLI and agent runtime, the project packages procedures, validation checks, and recovery guidance so agents can apply research and implementation knowledge rather than just retrieve summaries. DisCo's runtime builds on the open-source earendil-works/pi agent toolkit.
The library is designed for Codex, Claude Code, Pi, and other coding agents. Its router narrows a request to relevant skills, then loads only the needed instructions. This is useful when a task depends on practical repository workflows or research procedures that would otherwise require agents to find and assemble them on their own.
Key Features
- Provides a library of repository skills organized as graphs, with
SKILL.mdfiles and optional reference material and scripts. - Routes requests through the library so an agent can load only the relevant skill branch.
- Includes DisCo Researcher workflows for using installed skills and Creator workflows for constructing and refreshing skills.
- Distills repository and paper knowledge into operational instructions, including validation and recovery guidance.
- Exports selected skills and a scoped router to compatible coding agents such as Codex and Claude Code.
- Documents source provenance and refresh requirements for repository skills.
- Reports results on four autonomous research benchmarks, with the complete technical report still forthcoming.
Use Cases
- ML engineers can use relevant repository skills to set up and compare inference systems under matching workload and hardware constraints.
- Researchers reproducing a paper can follow skill-provided procedures and checks while recording commands and results.
- Teams maintaining agent workflows can create or refresh skills as upstream repositories change.
- Developers using Codex, Claude Code, or Pi can import selected skills to give an agent focused operational guidance for a task.
What you need
Detected in the repository
- Python (main language on GitHub)
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install DisCo, install the skill library, then start Researcher mode:
npm install -g --ignore-scripts @arex-skill/disco
disco repo-skills install
disco
See the README and installation guide for runtime setup, model provider configuration, and agent imports.
Alternatives
- book-to-skill: book-to-skill turns books and structured documents into reference skills, while AREX-Skill distills repositories and papers into verified, executable skills.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| AREX-Skill | Python | Apache-2.0 | 336 | New |
| book-to-skill | Python | MIT | 36k | New |
Considerations
AREX-Skill is a new project with rapid recent development, so workflows and library contents may change as it matures. The README reports six contributors and recent release activity, but does not provide a complete technical report for its benchmark results. The repository is Apache-2.0 by default, while each skill's own license metadata is authoritative; the DisCo npm package has a separate MIT license. DisCo requires Node.js 22.19.0 or newer, and model provider configuration is needed to use it.
Found this useful?
Share it with someone who would like AREX-Skill.