claude_config (timholy) skill
A scientist-developer's Claude Code setup for Julia: 26 skills for packages and manuscripts, three subagents and a comment hook
- 76 ★ on the repository
- Last commit
Extended access
Why this level
- The hook and skills run commands and edit package files, and extra dependencies install into the global Julia environment
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/claude-config-timholyDetects 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 hook and skills run commands and edit package files, and extra dependencies install into the global Julia environment
This is third-party code. Review the repository files before installing.
What it does
claude_config is a public copy of a Julia developer and scientist's personal `~/.claude` folder. The author says it is not a shared resource but a collection for the lab and for anyone who wants to learn from it. The repo holds `CLAUDE.md` instructions, rules in `rules/`, hooks, three subagents, Julia code fragments and 26 skills. For scientific work there are `/new-analysis-plan` and `/new-analysis-implement`: they split a long project into chunks and carry state through a handoff file. `/write-methods` helps understand an analysis the agent ran in a session, `/writing-manuscripts` and `/polish-manuscripts` handle manuscript style and consistency, and `/manuscript-benchmarks` builds a benchmark suite for a paper on a new method with result provenance. For Julia packages there are `/fix-issue`, `/update-compat` and `/profile-performance`, while the `/freshen-*` series polishes docstrings, docs, test coverage, Aqua, ExplicitImports, Runic and `.gitignore`. Three review skills, `/review-design`, `/review-api` and `/review-integration`, are complemented by `/review-implement`, which applies agreed changes in chunks. The `no-plan-comments.py` hook reminds the agent after every edit to strip plans and history from code comments.
Who it is for. Julia package developers and scientists who run data analysis and write papers with Claude Code and want working skills to learn from.
Good fit when
- A Julia package gets a new major dependency version and `compat` needs updating with `/update-compat`
- Package docstrings and docs need one style with `/freshen-docstrings` and `/freshen-docs`
- A multi-week data analysis is broken into stages with state passed between sessions
- A manuscript needs a consistency check of terms and numbers before submission
- Method performance benchmarks go into a paper and are rerun many times
Not a fit when
- You do not use Julia: the package skills serve only as inspiration
- You need a maintained shared resource: the author warns that issues and pull requests may be ignored
- You want one-command install: the files are copied into `~/.claude` by hand
Example request
Use /update-compat for my package: find dependencies with a new major version, update `compat` and run the tests.Limitations
The repo is meant as a personal `~/.claude` with many ignored files, so useful parts are copied into your own folder or a fork is cloned as `~/.claude` and the rest filled in. Hooks are registered in `settings.json`, which is not in the repo, and a ready example sits in `settings.example.json`. The Julia skills need Revise and TestEnv, and profiling also needs BenchmarkTools and FlameGraphs. Pandoc is recommended for manuscript export. The author does not promise to answer issues.
How to disable. Delete the copied skill, agent and hook folders from `~/.claude` and the `hooks` block from your own `settings.json`.
FAQ
How do I use the skills from this repo?
Copy the needed folders from `skills/` into your `~/.claude/skills/` and restart Claude Code. You do not have to install the whole folder.
Do I need Julia?
For the package skills yes, they target Julia. The science and manuscript skills, such as `/write-methods`, do not depend on the language.
What does the no-plan-comments hook do?
After every edit it reminds the agent if plans or work history leaked into code comments. It is registered in `settings.json`.
What is the freshen series?
A set of `/freshen-*` skills that each lift one aspect of package quality: docstrings, docs, coverage, Aqua, ExplicitImports, Runic and `.gitignore`.
How does package review work?
Three skills look from different angles: `/review-design` at the concept, `/review-api` at the interface, `/review-integration` at integration. Then `/review-implement` applies agreed changes in chunks.
Discussion
No questions or reviews yet.
Related
SkillsA Claude Code working configuration: a planner with per-milestone review, sub-agents and skills for analysis, refactoring and docs sync
Claude Code Blueprint: 12 agents, 19 skills, 16 hooks and 6 CLAUDE.md rules, copy what you need or install a preset
Claude Code plugins: plan, work and review commands, subagents, hooks and writing skills in a practical minimalism spirit.
An installer giving Claude Code rules for phases, decision logs, debugging and tests in Go, Rust, Python and six more languages.
A Claude Code plugin: 39 commands for PRs, review, commits and specs, nine expert roles and hooks, in eight language versions
A handbook and starter set for Claude Code: guides, CLAUDE.md templates, starter kits, hooks and the commit-helper plugin.
README in short
Covers the repo purpose, use by copying, the key components, science skills, coding and package maintenance skills for Julia and three review skills.