description

Initialize CWE in current project - creates workflow structure and checks plugin dependencies

allowed-tools

Write

Bash

Read

Glob

AskUserQuestion

Initialize CWE Project

Create the workflow structure for spec-driven development and ensure all recommended plugins are installed.

Step 1: Check Plugin Dependencies

CWE works best with these plugins installed:

Plugin	Purpose	Required
superpowers	TDD, debugging, planning, code review	Yes
serena	Semantic code analysis via LSP	Recommended
feature-dev	7-phase feature development	Recommended
frontend-design	Production-grade UI components	Optional
code-simplifier	Code cleanup and refactoring	Optional
claude-md-management	CLAUDE.md maintenance	Optional
plugin-dev	Plugin creation tools	Optional

Check installed plugins

Run this command to get installed plugins:

claude plugin list --json 2>/dev/null || echo '[]'

Compare with required plugins

Required: superpowers Recommended: serena, feature-dev Optional: frontend-design, code-simplifier, claude-md-management, plugin-dev

If plugins are missing

Use AskUserQuestion to ask the user:

Question: "Some recommended plugins are missing. Would you like to install them?"

Options:

"Install all missing" - Install all missing plugins
"Install required only" - Install only superpowers (if missing)
"Skip" - Continue without installing

Install missing plugins

For each plugin to install, run:

claude plugin install <plugin-name>

Show progress for each installation.

Step 1b: Check MCP Server Dependencies

CWE and its agents work best with these MCP servers:

MCP Server	Purpose	Level
playwright	Browser automation, E2E testing, screenshots	Recommended
context7	Up-to-date library docs via Context7	Recommended
github	GitHub API integration (PRs, issues, repos)	Recommended
filesystem	Direct filesystem access for agents	Optional
sequential-thinking	Step-by-step reasoning for complex tasks	Optional

Check installed MCP servers

Run this command to list currently configured MCP servers:

claude mcp list 2>/dev/null || echo 'No MCP servers configured'

Compare with recommended servers

Check which of the recommended servers are already configured.

If MCP servers are missing

Use AskUserQuestion to ask the user:

Question: "Some recommended MCP servers are missing. Install them?"

Options:

"Install all recommended" - Install playwright, context7, github
"Install all (recommended + optional)" - Install all 5 servers
"Let me pick" - Choose which to install
"Skip" - Continue without MCP servers

Detect platform

Before installing, detect the platform:

uname -s  # Linux, Darwin, MINGW/MSYS (Windows)

Install commands per platform

Linux / macOS (default):

claude mcp add playwright -- npx @playwright/mcp@latest --isolated --headless --no-sandbox
claude mcp add context7 -- npx @upstash/context7-mcp
claude mcp add github -- npx @modelcontextprotocol/server-github
claude mcp add filesystem -- npx @modelcontextprotocol/server-filesystem
claude mcp add sequential-thinking -- npx @modelcontextprotocol/server-sequential-thinking

Windows (MINGW/MSYS/WSL with Windows host):

# Playwright: remove --isolated --headless --no-sandbox flags (not supported on Windows)
claude mcp add playwright -- npx @playwright/mcp@latest
claude mcp add context7 -- npx @upstash/context7-mcp
claude mcp add github -- npx @modelcontextprotocol/server-github
claude mcp add filesystem -- npx @modelcontextprotocol/server-filesystem
claude mcp add sequential-thinking -- npx @modelcontextprotocol/server-sequential-thinking

Show progress for each installation. If a server fails to install, warn but continue with the rest.

Step 1c: Detect Serena Plugin (Memory Strategy)

If the serena plugin is installed, Serena provides its own memory system (write_memory, read_memory, list_memories). In that case, the CWE memory/ directory is redundant.

Check if serena is installed (from Step 1 results). If serena IS installed, use AskUserQuestion:

Question: "Serena plugin detected — it has its own memory system. Create CWE memory/ directory anyway?"

Options:

"Skip memory/ (use Serena)" — Do NOT create memory/ directory. Set SKIP_MEMORY=true.
"Create memory/ anyway" — Create memory/ as usual (both systems coexist).

Store the decision. If SKIP_MEMORY=true:

Do NOT create memory/ directory or any files within it
Do NOT add memory/ to .gitignore
Do NOT seed project-context.md or daily logs
Session hooks will gracefully skip memory operations (they already check for memory/ existence)

Step 1e: Install Statusline

CWE provides a live statusline for Claude Code showing context usage, rate-limit windows, and session time (no cost/currency tracking — keep it compact).

Check for existing statusline

test -f "$HOME/.claude/statusline.py" && echo "EXISTS" || echo "MISSING"

If EXISTS: Do NOT overwrite — the user may have a customized statusline. Skip to configuring Claude Code settings.

If MISSING: Copy the template:

cp "${CLAUDE_PLUGIN_ROOT}/templates/statusline.py" "$HOME/.claude/statusline.py"
chmod +x "$HOME/.claude/statusline.py"

Configure Claude Code settings

Add the statusline to the user's Claude Code settings.json (safe to run even if already configured):

claude config set --global status_line 'python3 ~/.claude/statusline.py'

If it fails or the user declines, skip — the statusline is optional but recommended.

Step 1f: Configure Web Research (optional)

CWE's /cwe:web-research command uses local self-hosted services for web search and scraping.

Ask the user with AskUserQuestion:

Question: "Hast du SearXNG und/oder Firecrawl lokal laufen?"
Header: "Web Research"
Options:
  1. "Ja, beides" - SearXNG + Firecrawl sind erreichbar
  2. "Nur SearXNG" - Nur Metasearch, Scraping via trafilatura
  3. "Nein, überspringen" - Web Research wird nicht konfiguriert (Recommended)

If "Ja, beides" or "Nur SearXNG":

Ask for URLs (defaults: SearXNG http://localhost:4000, Firecrawl http://localhost:3002).

Write or update .claude/cwe-settings.yml:

searxng_url: http://localhost:4000
firecrawl_url: http://localhost:3002

Quick connectivity test:

curl -s --max-time 3 "${SEARXNG_URL:-http://localhost:4000}/search?q=test&format=json" | python3 -c "import sys,json; d=json.load(sys.stdin); print(f'SearXNG OK — {len(d.get(\"results\",[]))} results')" 2>/dev/null || echo "SearXNG nicht erreichbar"

Step 1g: Configure Transcript Service (optional)

CWE's /cwe:transcript command supports Instagram/TikTok/Podcast URLs via a user-provided TScribe-compatible faster-whisper API. YouTube always works via tubetranscript fallback (no config needed).

Ask the user with AskUserQuestion:

Question: "Hast du einen TScribe-kompatiblen Whisper-Server fuer Instagram/TikTok/Podcast Transkripte?"
Header: "Transcript"
Options:
  1. "Ja, TScribe-URL eintragen" - User gibt eigene URL an
  2. "Nein, nur YouTube" - tubetranscript Fallback, skip TScribe

If "Ja":

Ask for URL (example: https://scribe.example.com or http://localhost:8000).

Write/update .claude/cwe-settings.yml:

tscribe_url: <user-url>

Step 1h: Configure Document Tools (optional)

Stirling PDF (for /cwe:pdf)

CWE's /cwe:pdf command reads PDFs by converting pages to images via Stirling PDF API (local, open-source). Without it, /cwe:pdf is disabled.

Question: "Hast du einen Stirling PDF Server fuer /cwe:pdf?"
Header: "PDF Reader"
Options:
  1. "Ja, URL eintragen" - Stirling ist erreichbar
  2. "Nein, spaeter" - /cwe:pdf wird nicht funktionieren bis konfiguriert
  3. "Docker starten?" - Start stirlingtools/stirling-pdf via Docker

If "Ja" — ask for URL (default: `http://localhost:8080`).

Write/update .claude/cwe-settings.yml:

stirling_pdf_url: http://localhost:8080

If "Docker starten" — suggest:

docker run -d --name stirling-pdf -p 8080:8080 stirlingtools/stirling-pdf:latest

Then write the URL as http://localhost:8080.

Step 1i: Configure Media Generation (optional)

CWE's media skills (/cwe:image, /cwe:video, /cwe:faceswap, /cwe:headswap, /cwe:upscale) require API keys.

Question: "Media-Generation-Tools einrichten?"
Header: "Media Tools"
Options:
  1. "Ja, API-Keys eintragen" - OpenRouter + MagicHour keys setup
  2. "Nur OpenRouter (Image only)" - Nur /cwe:image funktioniert
  3. "Nein, spaeter" - Skills installiert aber bis zur Konfiguration nicht nutzbar

If configuring keys:

Check if ${CLAUDE_PLUGIN_ROOT}/scripts/media-keys.sh already exists. If yes, ask before overwriting.

Ask for keys (without echoing to terminal — use masked input if possible):

OpenRouter (for /cwe:image): Sign up at https://openrouter.ai/settings/keys
MagicHour (for video/faceswap/etc.): Sign up at https://magichour.ai/

Write to ${CLAUDE_PLUGIN_ROOT}/scripts/media-keys.sh:

#!/usr/bin/env bash
# Media Tools API Keys — NEVER COMMIT (already in .gitignore)
export OPENROUTER_API_KEY="<user-input>"
export MAGICHOUR_API_KEY="<user-input>"

Set file permissions: chmod 600 scripts/media-keys.sh.

Step 1j: Configure Remotion (optional, for /cwe:motion)

Question: "Remotion-Projekt fuer /cwe:motion konfigurieren?"
Header: "Motion Graphics"
Options:
  1. "Bestehendes Projekt nutzen" - Pfad zu Remotion-Projekt angeben
  2. "Neues Projekt anlegen" - npx create-video@latest in ./remotion/
  3. "Nein, spaeter" - /cwe:motion wird nicht funktionieren bis konfiguriert

If existing:

remotion_project_dir: /path/to/remotion-project

If new:

npx create-video@latest remotion

Then:

remotion_project_dir: ./remotion

Step 1k: Configure Gitea Mirror (optional)

/cwe:gitea pushes to a private Gitea mirror. Credentials live in $HOME/.claude/cwe.local.md (not in cwe-settings.yml, to keep secrets out of project config).

Question: "Gitea-Mirror fuer /cwe:gitea einrichten?"
Header: "Gitea"
Options:
  1. "Ja, jetzt konfigurieren" - Guide user to create $HOME/.claude/cwe.local.md
  2. "Nein, spaeter" - Skip (Docs in commands/gitea.md)

If yes:

Show the Gitea-only block and ask the user to fill it in. Merge this block into $HOME/.claude/cwe.local.md — create the file with a YAML frontmatter section if missing, or append the gitea: key to the existing frontmatter. Do NOT add any bookstack: section here (that is Step 1l).

# $HOME/.claude/cwe.local.md
---
gitea:
  url: https://<your-gitea-host>
  user: <username>
  password: <token-or-app-password>
  ssh_host: <your-gitea-host>
  ssh_port: 22
---

Step 1l: Configure BookStack (optional)

BookStack is used by docs-oriented workflows for publishing notes/pages to a self-hosted knowledge base. Credentials live in the same $HOME/.claude/cwe.local.md as Gitea, under a separate bookstack: block.

Question: "BookStack fuer Notizen/Wissensbasis einrichten?"
Header: "BookStack"
Options:
  1. "Ja, jetzt konfigurieren" - Add bookstack: block to $HOME/.claude/cwe.local.md
  2. "Nein, spaeter" - Skip

If yes:

Show the BookStack-only block and ask the user to fill it in. Merge this block into $HOME/.claude/cwe.local.md under the same frontmatter section (add alongside gitea: if present; do NOT replace it).

# $HOME/.claude/cwe.local.md
---
bookstack:
  url: https://<your-bookstack-host>
  token_id: <token-id>
  token_secret: <token-secret>
---

Step 2: Check existing workflow setup

Check if workflow/ already exists:

If exists: Ask user if they want to reinitialize
If not: Proceed with creation

Step 3: Create structure

Create the following structure (skip memory/ section if SKIP_MEMORY=true):

workflow/
├── README.md              # Overview of workflow system
├── config.yml             # Project configuration
├── ideas.md               # Curated ideas backlog (per-project)
├── product/
│   ├── README.md          # What goes here
│   └── mission.md         # Product vision template
├── specs/
│   └── README.md          # How to write specs (folder-per-spec)
└── standards/
    └── README.md          # Project-specific standards (optional)

memory/                    # ONLY if SKIP_MEMORY is not true
├── MEMORY.md              # Index (200-line max, Hub-and-Spoke)
├── YYYY-MM-DD.md          # Daily logs (auto-created at session start/stop)
├── ideas.md               # Curated idea backlog
├── decisions.md           # Project ADRs
├── patterns.md            # Recognized work patterns
└── project-context.md     # Tech stack, priorities

docs/
├── README.md              # Project overview (auto-maintained)
├── ARCHITECTURE.md        # System design, component overview
├── API.md                 # Endpoint documentation (if applicable)
├── SETUP.md               # Installation, dev environment
├── DEVLOG.md              # Chronological developer journal
└── decisions/
    └── _template.md       # ADR template

VERSION                    # Single Source of Truth for version (semver)

Copy templates from ${CLAUDE_PLUGIN_ROOT}/templates/memory/ for all memory files (SKIP if SKIP_MEMORY=true). Copy templates from ${CLAUDE_PLUGIN_ROOT}/templates/docs/ for all docs files. Copy ${CLAUDE_PLUGIN_ROOT}/templates/docs/VERSION to project root.

Step 3b: Auto-Seed Project Memory

SKIP this entire step if SKIP_MEMORY=true (Serena handles memory).

After creating the memory structure, detect the project's tech stack and populate memory files.

Tech-Stack Detection

Check for these files in the project root:

File	Stack
`package.json`	Node.js — read `dependencies`/`devDependencies` for framework (React, Vue, Next.js, Express, etc.)
`tsconfig.json`	TypeScript (additional to package.json)
`Cargo.toml`	Rust — read `[dependencies]` for key crates
`go.mod`	Go — read module name
`pyproject.toml`	Python — read `[project.dependencies]` or `[tool.poetry.dependencies]`
`requirements.txt`	Python — list major packages
`composer.json`	PHP — read `require` for framework (Laravel, Symfony, etc.)
`Gemfile`	Ruby — read gems for framework (Rails, Sinatra, etc.)
`pom.xml`	Java — read dependencies
`build.gradle` / `build.gradle.kts`	Java/Kotlin — read dependencies
`Dockerfile`	Docker — note presence
`.github/workflows/`	GitHub Actions CI — note presence

Populate memory/project-context.md

Replace the (not configured) placeholders with detected values:

# Project Context

## Tech Stack
- Language: [detected language(s)]
- Framework: [detected framework(s)]
- Build: [detected build tool]
- CI: [detected if present]

## Current Phase
init

## Priorities
(not set — update after /cwe:start)

## Team
(not set)

Populate memory/MEMORY.md

Replace template placeholders:

{{project-name}} → actual directory name (basename "$PWD")
{{project-slug}} → lowercased, hyphenated directory name (basename "$PWD" | tr '[:upper:] ' '[:lower:]-')
Stack: (not configured) → detected tech stack summary
Phase: init → Phase: init (just initialized)

Placeholder Substitution (applies to all copied templates)

After copying templates, run these substitutions on the freshly-created files. Two kinds of placeholders exist:

Auto-filled by this command (must be substituted before presenting files to the user):

Placeholder	Value	Files
`{{PROJECT_NAME}}`	`basename "$PWD"`	docs/README.md, docs/SETUP.md, docs/API.md, docs/DEVLOG.md, workflow/config.yml
`{{project-name}}`	`basename "$PWD"`	memory/MEMORY.md
`{{project-slug}}`	`basename "$PWD" \| tr '[:upper:] ' '[:lower:]-'`	memory/MEMORY.md, memory/ideas.md, workflow/ideas.md
`{{DATE}}`	`date +%Y-%m-%d`	docs/DEVLOG.md, memory/.md

Example substitution block (run after templates are copied, skip memory/ lines if SKIP_MEMORY=true):

PROJECT_NAME="$(basename "$PWD")"
PROJECT_SLUG="$(basename "$PWD" | tr '[:upper:] ' '[:lower:]-')"
TODAY="$(date +%Y-%m-%d)"

# docs/
sed -i "s/{{PROJECT_NAME}}/${PROJECT_NAME}/g" docs/README.md docs/SETUP.md docs/API.md docs/DEVLOG.md 2>/dev/null || true
sed -i "s/{{DATE}}/${TODAY}/g" docs/DEVLOG.md 2>/dev/null || true

# workflow/
sed -i "s/{{PROJECT_NAME}}/${PROJECT_NAME}/g" workflow/config.yml 2>/dev/null || true
sed -i "s/{{project-slug}}/${PROJECT_SLUG}/g" workflow/ideas.md 2>/dev/null || true

# memory/ (skip if SKIP_MEMORY=true)
if [ "$SKIP_MEMORY" != "true" ]; then
  sed -i "s/{{project-name}}/${PROJECT_NAME}/g" memory/MEMORY.md 2>/dev/null || true
  sed -i "s/{{project-slug}}/${PROJECT_SLUG}/g" memory/MEMORY.md memory/ideas.md 2>/dev/null || true
fi

User-filled (left as natural-language hints like [e.g. npm install, ...] — the user edits these when populating their project). These are NOT placeholders to substitute; they are inline prompts. Do not run sed on them.

Examples intentionally left for the user to edit: install/dev/test/lint commands in README.md and SETUP.md, license, repo URL, API base URL, ADR title/deciders/number.

Create first daily log

Create memory/YYYY-MM-DD.md (using today's date) with:

# YYYY-MM-DD

## HH:MM — Project Initialized
- CWE initialized with /cwe:init
- Stack: [detected tech stack]
- Phase: init

VERSION file rules:

Plain text, single line, semver (e.g. 0.1.0)
ALL other version references (plugin.json, package.json, CHANGELOG, README) read from here
/cwe:devops release patch|minor|major bumps VERSION and cascades
Never hardcode version strings anywhere else

Step 3c: Patch .gitignore

CWE-generated files (memory/, workflow/, docs/) should NOT be committed to the target project's repo by default. After creating the structure, patch the project's .gitignore.

Check and append

Read the existing .gitignore (or create one if missing). Append the following block only if the CWE marker comment is not already present:

grep -q '# CWE (Code Workspace Engine)' .gitignore 2>/dev/null

If the marker is NOT found, append (omit memory/ line if SKIP_MEMORY=true):

# CWE (Code Workspace Engine)
memory/
workflow/
VERSION

Important:

Do NOT add docs/ — users may want to commit their project documentation
Do NOT add .claude/ — that's managed by Claude Code itself
If .gitignore doesn't exist, create it with the CWE block
Inform the user: "Added memory/ and workflow/ to .gitignore — CWE session data won't be committed."

File contents

workflow/README.md

# Workflow

This directory contains your project's workflow artifacts.

## Structure

- `config.yml` - Project configuration
- `ideas.md` - Ideas backlog for future development
- `product/` - Product vision, goals, roadmap
- `specs/` - Feature specifications
- `standards/` - Project-specific coding standards (optional)

## Quick Start

Run `/cwe:start` to begin guided development.

## Learn More

Run `/cwe:help` for full documentation.

workflow/config.yml

# CWE Project Configuration
version: "1.0"

project:
  name: "{{PROJECT_NAME}}"

workflow:
  phases:
    - plan      # Define product vision
    - spec      # Write feature specifications
    - tasks     # Break into implementable tasks
    - build     # Implement with agents
    - review    # Quality gates and verification

workflow/ideas.md

# Ideas Backlog

Curated ideas for this project. Raw observations in `~/.claude/cwe/ideas/<project-slug>.jsonl`.

## Status Legend

- **new** — Just captured, not yet reviewed
- **exploring** — Being discussed/developed
- **planned** — Approved for implementation
- **rejected** — Decided against

---

## Ideas

<!-- Ideas will be added here by the innovator agent -->

workflow/product/README.md

# Product

Define your product's vision, goals, and constraints here.

## Files

- `mission.md` - Core vision and goals (required)
- `roadmap.md` - Feature roadmap (optional)
- `constraints.md` - Technical/business constraints (optional)

workflow/product/mission.md

# Product Mission

## Vision

[What problem does this product solve? Who is it for?]

## Goals

- [ ] Primary goal
- [ ] Secondary goal

## Non-Goals

- What this product will NOT do

## Success Metrics

- How will you measure success?

workflow/specs/README.md

# Specifications

Feature specifications live here. Each feature gets its own folder.

## Structure

specs/ ├── YYYY-MM-DD-HHMM-feature-slug/ │ ├── plan.md # Implementation plan + task breakdown │ ├── shape.md # Scope, decisions, constraints │ ├── references.md # Similar code, patterns, prior art │ ├── standards.md # Standards snapshot at spec time │ └── visuals/ # Mockups, diagrams, screenshots


Folder naming is auto-generated: `YYYY-MM-DD-HHMM-<feature-slug>/`

## Creating a Spec

- Run `/cwe:start` → Spec Phase → "Shape-Spec Interview" (recommended)
- Or run `/cwe:architect shape` directly

workflow/standards/README.md

# Project Standards

Add project-specific coding standards here.

CWE loads built-in standards automatically via `.claude/rules/` with `paths` frontmatter.
8 domains: global, api, frontend, database, devops, testing, agent, documentation.

Use `/cwe:guide discover` to auto-discover patterns from your codebase.
Use `/cwe:guide index` to regenerate the standards index.

## Adding Custom Standards

Create `.claude/rules/your-standard.md` with `paths:` frontmatter for auto-loading.
Or add to `workflow/standards/` for project-specific conventions.

Step 3d: Multi-Terminal Development (optional)

Use AskUserQuestion:

Question: "Multi-Terminal Parallel Development einrichten?" Header: "Multi-Terminal"

Options:

"2 Terminals (Dev + QA)" — Preset A
"3 Terminals (Frontend + Backend + QA)" — Preset B
"4 Terminals (Frontend + Backend + Infra + QA)" — Preset C
"Custom (eigene Rollen)" — User defines roles
"Nein, später" — Skip

If "Nein, später": Skip to Step 4.

Preset Configuration

Preset	Branches	Handoff Files
A (2T)	`t1-dev`, `t2-qa`	`dev-qa.md`
B (3T)	`t1-frontend`, `t2-backend`, `t3-qa`	`frontend-backend.md`, `frontend-qa.md`, `backend-qa.md`
C (4T)	`t1-frontend`, `t2-backend`, `t3-infra`, `t4-qa`	`frontend-backend.md`, `frontend-qa.md`, `backend-qa.md`, `backend-infra.md`, `infra-qa.md`
D	User-defined	Generated from role pairs

For Custom (D): Ask the user

"Wie viele Terminals?" (2-6)
For each: "Rolle für Terminal {N}?" (e.g., api, worker, mobile)

Setup Steps (for any preset)

Create branches:

# For each terminal:
git branch t{N}-{role}

Create worktrees:

mkdir -p .trees
# For each terminal:
git worktree add .trees/t{N}-{role} t{N}-{role}

Create shared handoff directory:

mkdir -p shared/handoff

Generate handoff files from ${CLAUDE_PLUGIN_ROOT}/templates/multi-terminal/handoff-template.md: Replace {{SOURCE_ROLE}}, {{TARGET_ROLE}}, {{SOURCE_NUM}}, {{TARGET_NUM}} placeholders.
Copy handoff README:

cp "${CLAUDE_PLUGIN_ROOT}/templates/multi-terminal/README.md" shared/handoff/README.md

Generate terminal prompts from ${CLAUDE_PLUGIN_ROOT}/templates/multi-terminal/terminal-prompt.md:

mkdir -p terminal-prompts

Replace {{NUM}}, {{ROLE}}, {{ROLE_SLUG}}, {{AGENT_LIST}} placeholders.

Agent mapping for {{AGENT_LIST}}:

Frontend: builder, quality, architect, explainer
Backend: builder, devops, architect, security
QA: quality, security, researcher, ask
Infra: devops, security, architect, researcher
Dev (generic): builder, quality, architect, devops

Add to .gitignore:

.trees/

Initial commit on each branch:

# For each terminal branch:
git checkout t{N}-{role}
git add shared/ terminal-prompts/
git commit -m "chore: initialize multi-terminal structure for t{N}-{role}"
git checkout main

Show setup summary:

Multi-Terminal Development initialized!

Terminals:
  T1: {role} → .trees/t1-{role}/
  T2: {role} → .trees/t2-{role}/
  ...

To start a terminal:
  cd .trees/t{N}-{role} && claude

Commands:
  /cwe:autopilot      — Autonomous task loop
  /cwe:coordinate     — Team-lead coordination
  /cwe:check-handoff  — Check pending handoffs
  /cwe:handoff        — Send work to another terminal
  /cwe:qa-merge       — QA-verified merge to main

Step 4: Success message

Show completion summary:

CWE initialized successfully!

Plugins:
  superpowers (installed)
  serena (installed)
  feature-dev (installed)
  frontend-design (skipped)
  ...

Workflow structure created:
  workflow/
  ├── config.yml
  ├── ideas.md
  ├── product/mission.md
  └── specs/

Memory:
  memory/ created (CWE memory)
  — OR —
  memory/ skipped (using Serena memory)

Documentation structure created:
  docs/
  ├── README.md, ARCHITECTURE.md
  ├── API.md, SETUP.md, DEVLOG.md
  └── decisions/_template.md
  VERSION (0.1.0)

MCP servers configured:
  playwright (installed)
  context7 (installed)
  github (installed)
  filesystem (skipped)
  sequential-thinking (skipped)

Next steps:
1. Edit workflow/product/mission.md with your product vision
2. Run /cwe:start to begin guided development
3. Use /cwe:architect shape for your first feature spec

Adjust the plugin status based on what was actually installed/skipped.

FilesExpand file tree

init.md

Latest commit

History

init.md

File metadata and controls

Initialize CWE Project

Step 1: Check Plugin Dependencies

Check installed plugins

Compare with required plugins

If plugins are missing

Install missing plugins

Step 1b: Check MCP Server Dependencies

Check installed MCP servers

Compare with recommended servers

If MCP servers are missing

Detect platform

Install commands per platform

Step 1c: Detect Serena Plugin (Memory Strategy)

Step 1e: Install Statusline

Check for existing statusline

Configure Claude Code settings

Step 1f: Configure Web Research (optional)

If "Ja, beides" or "Nur SearXNG":

Quick connectivity test:

Step 1g: Configure Transcript Service (optional)

If "Ja":

Step 1h: Configure Document Tools (optional)

Stirling PDF (for /cwe:pdf)

If "Ja" — ask for URL (default: http://localhost:8080).

If "Docker starten" — suggest:

Step 1i: Configure Media Generation (optional)

If configuring keys:

Step 1j: Configure Remotion (optional, for /cwe:motion)

If existing:

If new:

Step 1k: Configure Gitea Mirror (optional)

If yes:

Step 1l: Configure BookStack (optional)

If yes:

Step 2: Check existing workflow setup

Step 3: Create structure

Step 3b: Auto-Seed Project Memory

Tech-Stack Detection

Populate memory/project-context.md

Populate memory/MEMORY.md

Placeholder Substitution (applies to all copied templates)

Create first daily log

Step 3c: Patch .gitignore

Check and append

File contents

workflow/README.md

workflow/config.yml

workflow/ideas.md

workflow/product/README.md

workflow/product/mission.md

workflow/specs/README.md

workflow/standards/README.md

Step 3d: Multi-Terminal Development (optional)

If "Nein, später": Skip to Step 4.

Preset Configuration

For Custom (D): Ask the user

Setup Steps (for any preset)

Step 4: Success message

If "Ja" — ask for URL (default: `http://localhost:8080`).