Claude Code Starter workflow
A Claude Code environment installer: CLAUDE.md, rules, skills, agents, hooks and two-layer project memory in one script
- 196 ★ on the repository
- Last commit
Extended access
Why this level
- The script creates files in the project, merges CLAUDE.md and settings.json and installs Claude Code hooks
- The orchestrator model lets the agent spin up workers and commit without confirming every step
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add workflows/claude-code-starterDetects 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 script creates files in the project, merges CLAUDE.md and settings.json and installs Claude Code hooks
- The orchestrator model lets the agent spin up workers and commit without confirming every step
This is third-party code. Review the repository files before installing.
What it does
Claude Code Starter is a single installer, init-project.sh. It puts a .claude/ structure into the project and builds a short CLAUDE.md instead of a mega-document. The script detects the scenario (new, existing or legacy project) and the project type: code, content or hybrid. Content projects (books, courses, knowledge bases, documents) get their own rules, skills and the writer, editor and content-reviewer agents. Project knowledge is split into two layers. Contracts are ARCHITECTURE.md, INVARIANTS.md and a methodology/ directory with a maturity ladder from draft to crystallized pattern. State is SNAPSHOT.md, BACKLOG.md and a dialogs/ directory where scripts/save-dialogs.sh and the /save-dialog skill keep the current session's JSONL before Claude Code deletes it by retention. The /start and /finish skills frame the beginning and end of a work session. Version 6.3.0 introduces an orchestrator model. The owner sets the goal and limits of authority and reviews the result, while Claude Code splits the work into streams, spins up workers, verifies and commits. An existing CLAUDE.md is merged additively, user sections survive, and on a hard conflict the proposal lands in .claude/CLAUDE.md.merge-proposal.md. An optional global layer in ~/.claude/ is installed by scripts/install-global.sh and adds the /setup-project skill.
Who it is for. Developers and authors of content projects on Claude Code whose CLAUDE.md has sprawled, whose rules are mixed with context and whose memory is lost between sessions.
Good fit when
- CLAUDE.md has grown past hundreds of lines and its rules need splitting into modules
- A new project is starting and needs a ready .claude/ structure with rules, skills and hooks
- A book, course or knowledge base is under way and needs the writer, editor and content-reviewer agents
- Claude Code sessions vanish by retention and dialogs must be kept inside the project
- An old project with scattered instructions must move to one scheme without losing its own sections
Not a fit when
- You need a single skill or hook: Starter installs a whole project structure
- Your main agent is not Claude Code: the .claude/ structure is built for it
- The project is not under git: git is required for installation
Example request
Install Claude Code Starter in this project, detect it as a content project and show which rules and agents were addedLimitations
- The installer is written in bash. It needs git and bash, and python3 is recommended for a safe merge of settings.json.
- The framework's service files can end up in shared git history. The README provides an explicit choice for how to treat them, so make it before publishing the repository.
- The orchestrator model lets the agent spin up workers and commit without confirming every step, so set the limits of authority up front.
- The documentation is in Russian and the latest commit is from 14 September 2026.
How to disable. Delete the .claude/ directory and the files the installer added (CLAUDE.md, SNAPSHOT.md, BACKLOG.md and others), or restore them from git. The global layer lives in ~/.claude/ and is removed by hand.
FAQ
How do I install Claude Code Starter?
Put init-project.sh in the project root and run `chmod +x init-project.sh`, then `./init-project.sh`. No flags are needed: the script detects whether the project is new, existing or legacy, and whether it is code, content or hybrid.
What happens to my CLAUDE.md?
The installer merges into it through scripts/lib/merge_claude_md.py: your sections survive and missing parts are added. On a hard conflict the install stops and writes a proposal to .claude/CLAUDE.md.merge-proposal.md.
How do I save Claude Code dialogs before they are deleted?
The scripts/save-dialogs.sh script and the /save-dialog skill copy the current session's JSONL into .claude/dialogs/. Saving also runs when you wrap up with /finish.
Can it be installed globally?
Yes, `bash /path/to/claude-code-starter/scripts/install-global.sh` puts the layer into ~/.claude/ and adds the /setup-project skill, so the framework is available in any folder.
Which project types does the installer detect?
Code, content and hybrid. Content projects get their own rules, skills and the writer, editor and content-reviewer agents.
Discussion
No questions or reviews yet.
Related
WorkflowsA starter kit for Claude Code: a CLAUDE.md memory-bank template with ready hooks, subagents and skills you copy into your project
Claude Code scaffold with 27 commands, 10 hooks and 25 skills installed globally to set common rules for every project.
A Claude Code working configuration: a planner with per-milestone review, sub-agents and skills for analysis, refactoring and docs sync
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 that generates and maintains CLAUDE.md: a 0 to 100 quality score, a 150-line cap and stale reference checks
22 skills for the full Claude Code project lifecycle: setup and task scope, code review, pre-push checks and session notes.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Claude Code Starter (this card) | no | 196 | |
| Claude Code Starter Kit | no | 342 | installs a ready global set of commands, hooks and skills, while here the installer detects the project type and adds two-layer memory |
| Harness Starter Kit | no | 116 | turns repeated agent mistakes into repository rules, while Starter builds the whole .claude/ structure and project memory |
README in short
The README, in Russian, explains why a control environment is needed, what is new in versions 6.0 to 6.3, the installer's requirements and how it detects the project type. It then describes three install options: a single file, a local checkout and a global layer with the /setup-project skill.