Skip to content
Back to search

Claude-Code-Development-Kit

Claude Code Workflow for beginners & intermediate users. Tutorial and Installer included

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/peterkrueck--claude-code-development-kit/badge.svg)](/listings/peterkrueck--claude-code-development-kit)

HTML

<a href="/listings/peterkrueck--claude-code-development-kit"><img src="/api/listings/peterkrueck--claude-code-development-kit/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

These steps run a script from the network that findsafeskills does not scan.

curl -fsSL https://raw.githubusercontent.com/peterkrueck/Claude-Code-Development-Kit/main/install.sh | bash

Or clone:

git clone https://github.com/peterkrueck/Claude-Code-Development-Kit.git
cd Claude-Code-Development-Kit
./setup.sh

The installer walks you through what to include — Full (recommended), Customize, or Minimal. Re-run anytime to add features.

**First session:**

cd your-project
claude
> /prime

Claude reads your docs, summarizes the project, and you're ready to work.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 176 lines · 2 findings marked

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Claude Code Development Kit
  2. **Keep Claude Code coherent across sessions.** A starter kit that maintains the context layer Claude actually reads.
  3. ## Why this exists
  4. A static `CLAUDE.md` decays within a few sessions. The context window fills up. Decisions you made yesterday get re-derived today because nothing kept the docs in sync with the code. Every fresh session starts from scratch, even when the answer is one `git log` away.
  5. This kit is the maintenance layer between you and Claude Code. Four focused doc files become the project's working memory. Slash commands keep them current. Hooks catch the things you'd forget. It's the setup you'd build yourself after a few weeks of using Claude Code, packaged so you start with it on day one.
  6. ## The loop
  7. ```
  8. /prime load project context at session start
  9. ↓
  10. work Claude writes the code
  11. ↓
  12. /review-work parallel sub-agents catch bugs + rule violations
  13. ↓
  14. /update-docs keep ai-context docs in sync (skipping trivial changes)
  15. ↓
  16. /merge verify and ship (standard branch or git worktree)
  17. ```
  18. `/verify` runs the app or its tests to confirm a change actually works (observed behavior, not just "it compiles") — the run-it-and-watch step `/review-work` points at.
  19. When Claude hits an architecture decision or a debugging dead end, ask for a **second opinion** — an independent review from a different model (OpenAI Codex by default, or Google Gemini).
  20. ## Who this is for
  21. You have a Claude Code subscription and want to hit the ground running. Maybe you're building an app, a side project, or starting something new. You want Claude to understand your project structure, review its own work, keep its own notes in sync — and not accidentally `git push --force` your main branch.
  22. This isn't a heavy framework. It's structured defaults that compound over time.
  23. ## Quick start
  24. ```bash
  25. curl -fsSL https://raw.githubusercontent.com/peterkrueck/Claude-Code-Development-Kit/main/install.sh | bash
    • Note: Pipes a downloaded script directly into a shell (curl | sh / wget | bash)
    • Line 38 on GitHub
  26. ```
  27. Or clone:
  28. ```bash
  29. git clone https://github.com/peterkrueck/Claude-Code-Development-Kit.git
  30. cd Claude-Code-Development-Kit
  31. ./setup.sh
  32. ```
  33. The installer walks you through what to include — Full (recommended), Customize, or Minimal. Re-run anytime to add features.
  34. **First session:**
  35. ```bash
  36. cd your-project
  37. claude
  38. > /prime
  39. ```
  40. Claude reads your docs, summarizes the project, and you're ready to work.
  41. ## What's in the box
  42. ### Core (always installed)
  43. | | |
  44. |---|---|
  45. | `CLAUDE.md` template | Your project's AI instruction set |
  46. | `/prime` command | Load core docs into context — tiered: a light prime by default (~4-6k tokens), `--full` for everything, `--deploy` to also load infra |
  47. | `/merge` command | Verify docs are current + tree is clean, then ship to main. Detects silent clean-merge breakage with a build/test pass. Works with standard branches and `git worktree`. |
  48. | `/verify` command | Run the app or its tests to confirm a change actually works — a cross-stack template you fill in for your stack |
  49. | `/update-docs` skill | Keep `docs/ai-context/` in sync after code changes — density-first, skips trivial diffs |
  50. | `context7-guidance` skill | Fetch current library/framework/API docs via Context7 instead of stale training data |
  51. | Doc scaffolding | Four focused files: `spec.md`, `project-structure.md`, `progress.md`, `deployment-infrastructure.md` |
  52. | Security scanner hook | Blocks API keys and secrets from leaking to MCP plugins |
  53. | Deny list + allowlist | Blocks `git push --force`, `rm -rf`, `sudo`, `git reset --hard`, etc.; auto-allows safe read-only shell utilities and doc-domain fetches |
  54. | Asset directories | `assets/` scaffolding for icons, logos, social, web |
  55. ### Quality gates (optional)
  56. | | |
  57. |---|---|
  58. | `/review-work` | Spawns parallel Claude sub-agents (Bug Hunter + Rules Auditor, plus an optional Architect) to review your uncommitted diff. Verifies every external-API claim against Context7 (unverified claims are discarded) and reports whether the change actually fulfills the task in `progress.md`. |
  59. | `/second-opinion` | Auto-triggers when Claude hits architecture decisions or debugging dead ends. Consults an external AI with a different model architecture — **OpenAI Codex by default**, **Google Gemini** on "ask Gemini", or "ask both" to triangulate. |
  60. | Review-on-stop hook | Three-stop advisory nudge, scoped to the files *this session* edited. First stop shows changes + review suggestions, second stop is a reminder, third stop lets you out. Never traps you. |
  61. `/second-opinion` needs the CLI for your chosen engine: [OpenAI Codex](https://github.com/openai/codex) and/or [Gemini CLI](https://github.com/google-gemini/gemini-cli). The installer asks which one(s) to wire up.
  62. ### Templates
  63. `AGENTS.md` (read natively by OpenAI Codex) and/or `GEMINI.md` (read natively by Gemini CLI) — the second-opinion consultant briefings — plus the four `docs/ai-context/` files and asset directory scaffolding.
  64. ### Optional add-ons
  65. - **Visual skills** — `/image-gen` (AI image generation via Gemini), `/image-edit` (crop/resize/rotate/mirror with content-bounds detection), `/bg-remove` (local background removal via rembg). Needs Python 3, Pillow, numpy; `/image-gen` needs a `GEMINI_API_KEY`; `/bg-remove` needs rembg.
  66. - **Deploy skill template** — A customizable deployment pipeline you fill in with your own commands. Follows: detect changes → run tests → deploy → verify → report.
  67. - **Audio notifications** — Plays a sound when Claude finishes a task or needs your input. macOS + Linux.
  68. ## What gets installed
  69. ```
  70. your-project/
  71. ├── .claude/
  72. │ ├── commands/
  73. │ │ ├── prime.md # /prime — load project context (tiered)
  74. │ │ ├── merge.md # /merge — verify and ship
  75. │ │ └── verify.md # /verify — run app/tests to confirm it works
  76. │ ├── hooks/
  77. │ │ ├── security-scan.sh # Blocks secrets from leaking to plugins
  78. │ │ ├── review-on-stop.sh # Advisory review nudge (if selected)
  79. │ │ ├── snapshot-baseline.sh # Session baseline capture (if selected)
  80. │ │ ├── track-file-touch.sh # Records this session's edits (if selected)
  81. │ │ ├── cleanup-session.sh # Clears per-session temp files (if selected)
  82. │ │ ├── notify.sh # Audio notifications (if selected)
  83. │ │ ├── config/
  84. │ │ │ ├── pipeline.json # Review-on-stop configuration
  85. │ │ │ └── sensitive-patterns.json # Security scan patterns
  86. │ │ └── sounds/
  87. │ │ ├── complete.wav
  88. │ │ └── input-needed.wav
  89. │ ├── skills/ # update-docs, context7-guidance (core) + review/visual/deploy
  90. │ └── settings.local.json # Permissions, hooks, deny list
  91. │
  92. ├── assets/ # Your visual assets
  93. │
  94. ├── docs/
  95. │ ├── ai-context/
  96. │ │ ├── spec.md # What the product does
  97. │ │ ├── project-structure.md # File tree and tech stack
  98. │ │ ├── progress.md # Roadmap and task tracking
  99. │ │ └── deployment-infrastructure.md # Hosting, secrets, CI/CD
  100. │ ├── legal/
  101. │ ├── business/
  102. │ ├── design-brand/
  103. │ └── open-issues/
  104. │
  105. ├── CLAUDE.md # Your project's AI rules
  106. ├── AGENTS.md # Codex consultant briefing (if selected)
  107. └── GEMINI.md # Gemini consultant briefing (if selected)
  108. ```
  109. ## Compatibility
  110. **Do I need an external CLI?**
  111. Only for `/second-opinion`. The installer asks which engine to wire up — [OpenAI Codex](https://github.com/openai/codex) (default), [Gemini CLI](https://github.com/google-gemini/gemini-cli), or both — and you only need the CLI for what you picked. The `/review-work` skill uses Claude sub-agents and works without either. Everything else works without them too.
  112. **Can I use this with Cursor / Windsurf / Codex?**
  113. The skills and hooks are Claude Code-specific. The documentation templates work with any AI tool. Skills should work with most AI coding tools that support the skill format.
  114. **Is this for large teams?**
  115. It works for teams, but it's designed for individual developers and small teams. If you're running complex multi-agent workflows at scale via the API, this probably isn't what you need.
  116. ## Upgrading from v2
  117. v3 is a major rewrite. Commands became skills, the 3-tier doc system became 4 focused files, Gemini moved from MCP to native CLI, Context7 is now a plugin.
  118. v2 is still available:
  119. ```bash
  120. curl -fsSL https://raw.githubusercontent.com/peterkrueck/Claude-Code-Development-Kit/v2/install.sh | bash
    • Note: Pipes a downloaded script directly into a shell (curl | sh / wget | bash)
    • Line 159 on GitHub
  121. ```
  122. See [CHANGELOG.md](CHANGELOG.md) for the full migration guide.
  123. ## Uninstalling
  124. Remove the installed files from your project:
  125. ```bash
  126. rm -rf .claude/ docs/ai-context/ assets/ CLAUDE.md AGENTS.md GEMINI.md
  127. ```
  128. The scaffolding directories (`docs/legal/`, `docs/business/`, etc.) are empty by default — remove them too if unused. Your code is never modified.
  129. ## License
  130. MIT — see [LICENSE](LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.