Onboarding wizard that customizes this Claude Code setup for your specific project. Collects project details through a structured intake questionnaire, then fills all CUSTOMIZE markers across CLAUDE.md and rule files automatically. Use when first adopting this setup for a new project ('customize this setup', 'configure for my project', 'fill in the CUSTOMIZE markers', 'onboard this setup'). Do NOT use for modifying hooks, agents, or other skills -- this only fills CUSTOMIZE markers in CLAUDE.md and rule files.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Onboarding wizard that customizes this Claude Code setup for your specific project. Collects project details through a structured intake questionnaire, then fills all CUSTOMIZE markers across CLAUDE.md and rule files automatically. Use when first adopting this setup for a new project ('customize this setup', 'configure for my project', 'fill in the CUSTOMIZE markers', 'onboard this setup'). Do NOT use for modifying hooks, agents, or other skills -- this only fills CUSTOMIZE markers in CLAUDE.md and rule files.
Do NOT modify hooks, agents, or other skills — this skill ONLY fills CUSTOMIZE markers
Do NOT invent information — if the user didn't provide something, ask for it
Keep it concise — CLAUDE.md is loaded on every prompt. Verbose sections waste tokens.
Match the style of existing content in each file when filling markers
What This Skill Does
This setup has <!-- CUSTOMIZE --> markers in CLAUDE.md and .claude/rules/*.md that need
project-specific details. This skill collects your project information ONCE, then fills in
ALL markers automatically.
See examples.md in this skill directory for a complete example of what good customization
looks like (the "Acme HR" reference project).
Task Tracking
Tasks survive context compacts — skipping this check causes lost progress and repeated work.
Before starting work, run TaskList to check if tasks already exist from a previous session or before a compact. If tasks exist:
Read existing tasks with TaskGet for each task ID
Find the first task with status pending or in_progress
Resume from that task — do NOT recreate the task list
If no tasks exist, create them after the first round of user questions:
Example task list:
Task 1: Collect project brief (all rounds)
Task 2: Read all CUSTOMIZE markers
Task 3: Fill CLAUDE.md
Task 4: Fill rule files
Task 5: Clean up markers
Task 6: Validate completeness
Mark each task in_progress when starting and completed when done.
Step 1: Collect Project Brief
Use AskUserQuestion to collect project details. Ask up to 4 questions at a time, then
follow up with remaining questions. Do NOT proceed until all critical information is gathered.
Round 1 — Core Identity
Ask these questions using AskUserQuestion:
Project name and description — "What is your project name and a one-line description?"
This is the most important file — Claude reads it on every conversation.
Marker: Project description (top of file)
Replace the <!-- CUSTOMIZE: Replace with a brief description --> comment with a one-line
description. Example: Acme SaaS — Next.js App Router, Supabase, TypeScript.
Remove the entire HTML comment block (the template instructions at the top of the file).
Marker: Critical Rules
Add any framework-specific critical rules from the intake. Examples:
If using a Server Action wrapper: "Skipping {wrapper} means unauthenticated data can reach the database."
If using a specific logger: "Use getLogger() from {path} instead of console.log."
If monorepo with upstream: "When merging upstream, propagate infrastructure changes to all product apps."
Marker: Monorepo
If monorepo: Add a table with app names, ports, and purposes
If single-app: Remove the ## Monorepo section entirely
Marker: Commands
Add all commands from the intake in a code block. Group by category (dev, test, build, deploy).
Marker: Architecture
Describe: multi-tenant model, data fetching patterns, auth, RLS approach, any special systems.
Marker: Verification
Add the project's verify/check commands (typecheck, lint, test).
Step 4: Fill Rule Files
For each rule file, replace <!-- CUSTOMIZE --> markers with project-specific content.
git-workflow.md
Remotes table (origin URL, upstream if applicable)
Branch strategy (adjust the table)
Verify command
CI pipeline table
Upstream merge process (if forked from template, otherwise remove section)
database.md
Migration/typegen command paths
RLS helper functions (list any existing helpers, or keep the example)
Server Action wrapper (replace manual pattern with framework wrapper if applicable)
Supabase client import paths (update the table)
Auth/account model (adjust or remove multi-tenant section)
coding-style.md
Server Action wrapper reference (match patterns.md)
security.md
Auth wrapper documentation
ui-components.md
Component library name and import path
forms.md
Server Action wrapper
Form component library imports
testing.md
Test directory structure
Test commands
i18n.md
i18n setup (library, config)
Translation namespaces
If no i18n: Delete the entire file (or leave as-is for future use)
admin.md
Admin verification pattern
pages-and-layouts.md
Loading component
Workspace providers
Feature flags (if applicable)
route-handlers.md
Route handler wrapper (if framework provides one)
Step 5: Clean Up Markers
After filling all markers:
Remove all <!-- CUSTOMIZE --> comments — they served as prompts and should be
deleted after customization. The content they prompted for is now filled in.
Remove the template instruction block at the top of CLAUDE.md (the HTML comment
with "SETUP INSTRUCTIONS").
Update file headers — Change generic titles to project-specific ones:
# Patterns — Next.js Supabase TypeScript could become
# Patterns — Acme Next.js Supabase
Customization complete for {project-name}!
Files modified:
- CLAUDE.md
- .claude/rules/git-workflow.md
- .claude/rules/database.md
- ... (list all modified files)
Remaining markers: {count} (0 = fully customized)
Next steps:
1. Review the changes: git diff
2. Test the setup: start a new Claude Code session and verify context
3. Optional: Add project-specific validators to post_tool_use.py
Resuming After Context Compact
If you notice context was compacted or you're unsure of current progress:
Run TaskList to see all tasks and their status
Find the in_progress task — that's where you were
Run TaskGet {id} on that task to read full details
Continue from that task — don't restart from the beginning
Tasks persist across compacts. The task list is your source of truth for progress, not your memory.
Pattern for every work session:
TaskList → find in_progress or first pending → TaskGet → continue work → TaskUpdate (completed) → next task
Troubleshooting
Markers remain after Step 6 validation
The most common cause is a marker inside an HTML comment that looks like content.
Read the file, find the <!-- CUSTOMIZE comment, and either fill it or delete it.
If the marker needs information you don't have, use AskUserQuestion to ask.
User unsure about commands or paths
Offer to read their package.json for commands: Read package.json and extract scripts.
For monorepo structures, Glob "apps/*/package.json" reveals app layout.
i18n section not applicable
If the project doesn't use i18n, you can either delete .claude/rules/i18n.md entirely
or leave it with a note at the top: "i18n is not currently used in this project."
User provides minimal answers
Ask follow-up questions. A one-word answer like "pnpm" is fine for package manager, but
"yes" for monorepo needs follow-up: "What are your app names, ports, and purposes?"
