L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Développement · SKILL.md

init

init

Interactive setup wizard for claudekit. Scaffolds rules, hooks, and MCP server configs into the user's project. Run /claudekit:init to configure. Use when setting up a new project with claudekit or reconfiguring an existing one.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: init description: > Interactive setup wizard for claudekit. Scaffolds rules, hooks, and MCP server configs into the user's project. Run /claudekit:init to configure. Use when setting up a new project with claudekit or reconfiguring an existing one. user-invocable: true argument-hint: "[--all] to skip prompts and install everything"

Claudekit Init

Interactive setup wizard that scaffolds project-level configuration files into the user's .claude/ directory.

Output styles ship with the plugin and are auto-discovered by Claude Code (no init step needed for them — see output-styles/ at the plugin root).

What It Generates

CategoryFilesLocation
Rulesapi.md, frontend.md, migrations.md, security.md, testing.md.claude/rules/
Hooksauto-format, block-dangerous-commands, detect-secrets, guard-sensitive-files, notify.claude/hooks/ + settings.local.json
MCP Serverscontext7, sequential, playwright, memory, filesystem.mcp.json

Wizard Flow

When invoked, ask the user ONE question at a time:

Step 1: Rules

"Which rules do you want to install?"

  • a) All rules (api, frontend, migrations, security, testing)
  • b) Let me pick individually
  • c) Skip rules

If (b), list each rule with a one-line description and let user select:

  • api.md — REST API design conventions (naming, versioning, error responses)
  • frontend.md — React/Next.js component patterns and file organization
  • migrations.md — Database migration safety rules (backward compatibility, rollback)
  • security.md — OWASP-aligned security rules (no hardcoded secrets, parameterized queries)
  • testing.md — Test naming, coverage thresholds, mocking conventions

For each selected rule, read the template from ${CLAUDE_PLUGIN_ROOT}/skills/init/templates/rules/<name>.md and write it to .claude/rules/<name>.md.

Step 2: Hooks

"Which hooks do you want to install?"

  • a) Auto-format (runs linter after Write/Edit)
  • b) Block dangerous commands (prevents rm -rf /, force push main, etc.)
  • c) Detect secrets (blocks writes containing API keys, tokens, private key blocks)
  • d) Guard sensitive files (blocks edits to .env files, key material, credential dotfiles)
  • e) Notifications (desktop notifications on completion)
  • f) All of the above
  • g) Skip hooks

For each selected hook:

  1. Read the hook metadata from ${CLAUDE_PLUGIN_ROOT}/skills/init/templates/hooks.json
  2. Copy the hook script from ${CLAUDE_PLUGIN_ROOT}/scripts/<script>.cjs to .claude/hooks/<script>.cjs
  3. Merge the hook entry into .claude/settings.local.json (create if it doesn't exist)

Hook entry format for settings.local.json:

{
  "hooks": {
    "<event>": [
      {
        "matcher": "<matcher>",
        "hooks": [
          {
            "type": "command",
            "command": "node .claude/hooks/<script>.cjs"
          }
        ]
      }
    ]
  }
}

If settings.local.json already has a hooks key, merge new entries into the existing structure — do not overwrite.

Step 3: MCP Servers

"Which MCP servers do you want to configure?"

  • a) Context7 (library documentation lookup)
  • b) Sequential Thinking (multi-step reasoning)
  • c) Playwright (browser automation)
  • d) Memory (persistent knowledge graph)
  • e) Filesystem (secure file operations)
  • f) All of the above
  • g) Skip MCP setup

For each selected server:

  1. Read the server config from ${CLAUDE_PLUGIN_ROOT}/skills/init/templates/mcp-servers.json
  2. Detect platform: check if process.platform === "win32" or use Bash uname to determine OS
  3. Select the correct config (win32 or posix key)
  4. Merge into the project's .mcp.json (create with {"mcpServers": {}} if it doesn't exist)

Step 4: Summary

Print a summary table of everything installed:

Claudekit setup complete!

  Rules:   5 installed → .claude/rules/
  Hooks:   5 installed → .claude/hooks/ + settings.local.json
  MCP:     5 configured → .mcp.json

Next steps:
  - Skills available as /claudekit:<name> (15 total)
  - Agents available as claudekit:<name> (8 specialists)
  - Output styles available via /config (5 shipped: Brainstorm, Deep Research,
    Implementation, Review, Token Efficient)

--all Flag

If $ARGUMENTS contains --all, skip all prompts and install everything:

  • All 5 rules
  • All 5 hooks
  • All 5 MCP servers

Important Rules

  • NEVER overwrite existing files without asking. If a target file already exists, ask: "[filename] already exists. Overwrite? (y/n)"
  • Create directories as needed. If .claude/rules/ doesn't exist, create it before writing files.
  • For hooks, always use settings.local.json (not settings.json) — local is gitignored so hook config stays personal.
  • Use ${CLAUDE_PLUGIN_ROOT} to reference template files within the plugin.
  • Platform detection for MCP: Windows uses cmd /c npx, macOS/Linux uses npx directly.
  • Output styles are NOT scaffolded by init. They ship with the plugin at output-styles/ and are auto-discovered. Users switch them via /config or by setting outputStyle in .claude/settings.local.json.
Skills enregistrées