These are intended for agents but they can also be read by humans to understand philosphy and best practice.
See:
- https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview
- https://github.com/ai4curation/curation-skills
Currently there is only one, for repo best practices:
Everyone should read the skills in this repo and check they are in agreement - PRs always welcome!
In Claude Code you can install these via the marketplace
/plugin marketplace add berkeleybop/bbop-skills
Or just copy into your own skills folder.
But it's as important people read and understand the skills as it is that their agents use them
All group members are expected to develop core AI capabilities and literacy, and to use judgment in applying them. There are favored tools such as Claude Code for the terminal and Claude Code on the web for non-terminal use, but no mandate to use these.
The existence of these tools does mean that expectations of things like basic repo hygiene and best practices as laid out in github-repo-skill/ are now much higher, given it is very easy for coding tools to do a good job on this, even with legacy repos.
While we generally want to be forward in using AI, we should be cogniscent of context, see for example:
We used to maintain a github repo here with best practice:
https://berkeleybop.org/best_practice/
A lot of this has become stale, and many parts of this will gradually be migrated to skills, either in this repo, or in a repo such as the ai4curation skills repo
The LICENSE.txt is CC-0. Feel free to reuse, adapt. Attributions welcome but not required.