Julius Skills
Eight personal skills by Julius Brussee: terse replies, interfaces, plan grilling, a task loop, review, text cleanup, context checks
- 175 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skills are instructions and reference files and run no commands themselves
- On request `loop-factory` installs a CLI from a separate repository into the project
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/julius-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills are instructions and reference files and run no commands themselves
- On request `loop-factory` installs a CLI from a separate repository into the project
This is third-party code. Review the repository files before installing.
What it does
Julius Skills is a collection of eight skills the author keeps for himself and published as one pack. The first, `caveman`, shortens agent replies: it drops filler and pleasantries but keeps commands, code and error text exact. `interface-kit` gathers rules for accessibility, typography, grids, color and states, while `grill-me` first finds out how well you know the topic and only then asks one question at a time, raising the pressure. `loop-factory` keeps tasks as markdown specs in inbox, active and archive folders and closes them only after a review. `junior-to-senior` treats the agent's plan as a junior's work and rewrites it after studying the code and fresh practices from the web. `deslopify` looks for AI-writing tells with a catalog of regular expressions, rewrites and rescans. `context-canary` puts a control line at the start of each reply whose disappearance shows that the agent is losing its instructions, and describes a recovery routine. `last-20-percent` takes a finished solution through the last 20 percent: the usage scene, first run and defaults.
Who it is for. Developers who work with Claude Code or Codex all day and want to tune the agent's manner: answer more briefly, check plans more strictly and not lose instructions in long sessions.
Good fit when
- Agent replies are bloated and you want shorter ones without losing commands and code
- A plan looks vague in the hard parts and needs a rework the way a senior engineer would do it
- The idea is still raw and one question at a time with answer options will help work it down to the forks
- A long session starts forgetting project rules and you need a signal when it happens
- A draft with AI-writing tells has to be brought to a human text in the right register
Not a fit when
- You need one pre-vetted skill with a wide audience: these skills are personal and fit the author's habits
- You want the full Caveman setup with its own installer: the author keeps it in a separate Caveman repository
- You need a full task-loop engine with an interface: this is only a skill, while the CLI sits in a separate Loop-Factory repository
- You need cleanup of Russian text: `deslopify` is built around English tells
Example request
Switch on caveman and explain why the build fails only in CI, then review my fix plan with junior-to-seniorLimitations
The repository is meant to be personal, so skill descriptions are short and not every skill has reference files. `loop-factory` needs a separate repository with the CLI and adapters, and the canonical `caveman` installer is an `install.sh` script from another repository. `deslopify` and `context-canary` are described only for English text. The GitHub description is empty, so the purpose of the pack is visible only from the README.
How to disable. Delete the folders of the skills you want from the agent's skills directory, for example `.claude/skills/caveman`. If `loop-factory` created spec folders in a project, delete them by hand.
FAQ
How do I install Julius Skills in Claude Code?
Run `npx skills@latest add JuliusBrussee/skills` and mark skills and agents in the interactive menu. The installer supports Claude Code, Codex, Gemini, Cursor, Windsurf, Cline and Copilot.
How does `caveman` here differ from the separate Caveman repository?
The pack contains the skill itself with the terse reply style. The canonical installer lives in the Caveman repository and is run with its `install.sh`.
How does `grill-me` work?
The skill first asks how much you know about the topic and how much pressure you want, then asks one question at a time with a suggested answer. You can make it softer or harder with a command in the chat.
What is the context canary?
A control line at the start of each reply: a name, a turn number and a context self-check. If the line disappears, instructions were lost after compaction or truncation, and the skill advises saving state and starting a fresh session.
Can I take just one of the eight skills?
Yes, the installer shows a list and each skill sits in its own `skills/` folder with its own `SKILL.md`. Only `loop-factory` has a dependency, a CLI from a separate repository.
Discussion
No questions or reviews yet.
Related
Skills14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
Builder Skills
Skills for agents
A set of small skills from Builder.io for agent work habits: visual plans and diff recaps, docs-first checks, cross-agent review and clean status updates
Five slash commands for an agent: a plan before code, memory across sessions, review, failure triage and one UI registry
Dimillian Skills
Skills
A set of 16 engineering skills for agents: Apple platform work, macOS packaging, React performance, code review and refactoring
A set of independent skills against drift in large codebases: hard-cut, root-cause tracing, lean testing and CI right-sizing.
my-skills by akitaonrails
akitaonrails my-skills
Fabio Akita's personal set of 24 skills for Claude Code, Codex and opencode: releases, PR audits, security and fact-checking
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Julius Skills (this card) | no | 175 | |
| Caveman | no | 117,983 | a separate repository of the reply-compression skill with its own installer |
| grill-me: plan interview skill | no | 1,403 | a standalone plan-grilling skill with no pressure calibration by knowledge level |
| Humanizer | no | 65,341 | removes AI-writing tells by the Wikipedia list, while deslopify loops over a regular-expression catalog |
README in short
The README briefly names the eight skills and the pack's principle: fewer words, careful interfaces, calibrated plan grilling and a disciplined task loop. It then gives install commands, explains three common ways agent output fails and describes each skill with a list of when to use it.