Claude Code Toolkit (robertguss) skill
Claude Code Toolkit
A set of Claude Code skills and hooks: brainstorming, worker-led project leadership, a wiki, and a book pipeline.
- 135 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Hooks run formatters and a TypeScript check automatically after every file edit.
- The lead skill delegates tasks to external worker agents in Herdr.
Install
- Works in
Manual install
python build.py brainstormPackages the chosen skill into `dist/brainstorm.skill` for manual upload to Claude.ai, under Settings, Skills.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks run formatters and a TypeScript check automatically after every file edit.
- The lead skill delegates tasks to external worker agents in Herdr.
This is third-party code. Review the repository files before installing.
What it does
Claude Code Toolkit bundles eight skills, three hooks and three templates into one repository. The brainstorm skill runs a multi-session idea exploration through a method catalog, lead hands tasks to fresh workers in Herdr and verifies them with its own probes, and project-wiki builds a Karpathy-style project wiki and publishes it with Quartz. handoff prepares a session-continuity document, while ebook-factory, non-fiction-book-factory and writing cover the chain from idea to finished book while keeping the author's voice. The hooks run formatting after edits (`auto-format`), write a change summary when a session ends (`change-summary`), and tell Claude what to preserve during context compaction (`compaction`). The `HUMAN.md` and `CLAUDE.md` templates set how Claude Code sees the user and the project.
Who it is for. Claude Code users running long multi-session work, writing non-fiction books, or wanting auto-formatting and a change summary without hand-rolling the hooks.
Good fit when
- An idea exists only in rough form and needs a brainstorming partner with several structured methods, not one pass.
- Several Herdr worker agents run a project, and results need independent probes rather than trusting the report.
- A session needs to end but its context matters for the next one, so a handoff document beats a mental recap.
- File edits happen often, and automatic ruff, goimports or prettier formatting beats hand-rolling the hook.
Not a fit when
- The task is one file and a couple of edits: eight skills and templates are overkill for that.
- There is no Herdr worker infrastructure, and the lead skill is built around it specifically.
- A finished book with no further work is expected: the ebook-factory and non-fiction-book-factory pipeline gives architecture and voice, not a final text without your own editing.
Example request
Run a brainstorm on this new product idea using several methods, then put together a handoff document for the next session.Limitations
The lead skill assumes Herdr worker infrastructure; without it, handing tasks to fresh agents only works partially. For Claude.ai (web, mobile, desktop) each skill has to be packaged separately with `python build.py <name>` and the resulting `.skill` file uploaded by hand — there is no bulk upload. The `change-summary` hook runs a TypeScript check, so in non-TypeScript projects it mostly just gives a change summary. MIT licensed; the author supports it through issues and pull requests in their own time.
How to disable. Remove the reference to that skill from the project's `CLAUDE.md`, or delete its folder. Hooks are disabled by removing the matching block from `~/.claude/settings.json` and the script file from `~/.claude/hooks/`.
FAQ
How do I wire up one skill from the set in Claude Code?
Clone the repository and add a line in `CLAUDE.md` pointing to the needed `SKILL.md`, e.g. for brainstorm: "read and follow /path/claude-code-toolkit/skills/brainstorm/SKILL.md".
How do I upload a skill to Claude.ai instead of Claude Code?
Package it with `python build.py brainstorm` (swap in the skill's name), get the file from `dist/`, then upload it manually in Claude.ai, under Settings, Skills.
What does the auto-format hook do?
It fires on the PostToolUse event after Edit, MultiEdit or Write and runs the project's formatter — ruff, goimports or prettier, depending on the file type.
Does every skill need Herdr?
No, only the lead skill needs Herdr to hand out tasks to workers. Brainstorm, handoff, project-wiki and the book skills work without it.
How does HUMAN.md differ from CLAUDE.md in the templates?
HUMAN.md describes the user to Claude so the agent remembers who it is working with. CLAUDE.md holds global instructions for every project, meant to be copied and edited to taste.
Discussion
No questions or reviews yet.
Related
SkillsAbout a hundred skills for Claude Code: meetings, research, publishing, TDD, Telegram, email and personal analytics
A skill that keeps project memory in plain Markdown files: session logs, a wiki, rules, a to-do list and a note on where to resume
A Claude Code plugin that builds a knowledge system of markdown files from a conversation with you: folder structure, note templates, a processing pipeline and
Claude Code skills for product and UX work: website UX audit, requirement docs, a PMF cycle, text clustering and ADRs
Second Brain Skills
Claude Code Second Brain Skills
A set of Claude Code skills for knowledge work: brand and voice, an MCP client, PPTX slides, Remotion video, SOPs and a skill creator
A Claude Code plugin that catches your corrections in chat and, after review, writes them into CLAUDE.md and AGENTS.md
README in short
The author splits the set into skills, hooks and templates, showing for each one how to wire it up via CLAUDE.md or install the hook. A separate GitHub Pages site covers each skill in more depth than the README.