create-claudify 1.0.0

package/templates/.claude/skills/deploy/SKILL.md

@@ -0,0 +1,40 @@
+# Deploy Skill
+
+Deployment procedures for shipping code to production.
+
+## When to Load
+
+- Deploying to staging or production
+- Configuring CI/CD pipelines
+- Troubleshooting deployment failures
+
+## Pre-Deploy Checklist
+
+1. All tests pass: `{{TEST_COMMAND}}`
+2. Build succeeds: `{{BUILD_COMMAND}}`
+3. No TypeScript errors: `{{TYPECHECK_COMMAND}}`
+4. Lint clean: `{{LINT_COMMAND}}`
+5. Environment variables configured for target environment
+6. Changes committed and pushed
+
+## Deployment Targets
+
+### {{PLATFORM_NAME}}
+- **URL:** {{DEPLOY_URL}}
+- **Deploy command:** {{DEPLOY_COMMAND}}
+- **Environment:** {{ENV_VARS_NEEDED}}
+
+## Post-Deploy Verification
+
+1. Check the deployed URL loads correctly
+2. Test critical user flows
+3. Check for console errors
+4. Verify API endpoints respond
+5. Log the deployment in daily note
+
+## Rollback
+
+If something goes wrong:
+1. {{ROLLBACK_PROCEDURE}}
+2. Log the incident in `.claude/logs/incidents.md`
+3. Nominate the failure to knowledge-base

package/templates/.claude/skills/monitor/SKILL.md

@@ -0,0 +1,44 @@
+# Monitor Skill
+
+System and application health monitoring procedures.
+
+## When to Load
+
+- Checking if services are running correctly
+- Investigating errors or performance issues
+- Regular health checks
+
+## Health Checks
+
+### Application
+- Is the app accessible at the expected URL?
+- Are API endpoints responding with expected status codes?
+- Are background jobs/cron tasks running?
+
+### Infrastructure
+- Server resource usage (CPU, memory, disk)
+- Database connection pool status
+- External service dependencies (APIs, CDNs)
+
+### Code Quality
+- Test suite passing?
+- Build succeeding?
+- Linting clean?
+- Dependency vulnerabilities (`npm audit`)?
+
+## Alert Thresholds
+
+| Metric | Warning | Critical |
+|---|---|---|
+| Response time | > 1s | > 5s |
+| Error rate | > 1% | > 5% |
+| Disk usage | > 80% | > 95% |
+
+## Investigation Procedure
+
+1. **Identify**: What symptom was observed?
+2. **Scope**: Is it affecting all users or specific conditions?
+3. **Correlate**: Did anything change recently? (deploys, config, dependencies)
+4. **Diagnose**: Check logs, metrics, and error traces
+5. **Fix or escalate**: Apply fix if clear, otherwise escalate with findings
+6. **Document**: Log the incident and root cause for knowledge-base

package/templates/.claude/skills/research/SKILL.md

@@ -0,0 +1,42 @@
+# Research Skill
+
+Structured approach to investigating a topic using web search and subagents.
+
+## When to Load
+
+- User asks to research, investigate, or explore a topic
+- Need to gather information from multiple sources before acting
+
+## Procedure
+
+1. **Define the question**: What specifically do we need to know?
+2. **Identify sources**: Where would reliable answers come from?
+3. **Parallel search**: Use Task tool with multiple subagents for independent queries
+4. **Cross-reference**: Compare findings across sources for consistency
+5. **Synthesize**: Distill into actionable conclusions with source attribution
+6. **Store**: Key findings go to daily note; durable facts nominated to knowledge-base
+
+## Output Format
+
+```markdown
+## Research: [Topic]
+
+**Question:** [What we investigated]
+**Sources:** [Where we looked]
+
+### Findings
+- [Finding 1] — [Source]
+- [Finding 2] — [Source]
+
+### Conclusions
+- [Actionable takeaway]
+
+### Confidence: HIGH | MEDIUM | LOW
+```
+
+## Tips
+
+- Prefer official docs over blog posts
+- Cross-reference numbers across sources — if they disagree, flag it
+- Date-stamp findings — information decays
+- If confidence is LOW, say so explicitly instead of guessing

package/templates/.mcp.json

@@ -0,0 +1,14 @@
+{
+  "mcpServers": {
+    "context7": {
+      "command": "npx",
+      "args": ["-y", "@upstash/context7-mcp@latest"],
+      "description": "Live library documentation (Next.js, React, any npm package)"
+    },
+    "memory": {
+      "command": "npx",
+      "args": ["-y", "@modelcontextprotocol/server-memory"],
+      "description": "Persistent knowledge graph for cross-session facts"
+    }
+  }
+}

package/templates/CLAUDE.local.md

@@ -0,0 +1,18 @@
+# Personal Overrides
+
+This file is gitignored — your personal preferences that don't affect the team.
+
+## My Preferences
+
+- [Your preferred coding style differences]
+- [Your editor/IDE specifics]
+- [Your local environment quirks]
+
+## My API Keys & Services
+
+- [Note which services you have configured locally]
+- [Never put actual keys here — use .env files]
+
+## My Workflow
+
+- [Any personal workflow adjustments]

package/templates/CLAUDE.md

@@ -0,0 +1,91 @@
+# {{PROJECT_NAME}}
+
+> Production-grade Claude Code project instructions.
+> Generated by [Claudify](https://claudify.dev). Keep under 150 lines.
+
+## Quick Start
+
+- Run `/start` to begin a work session
+- Run `/sync` to refresh context mid-session
+- Run `/clear` to safely flush and resume when context feels heavy
+- Run `/wrap-up` to close out a session
+- Run `/audit` to review recent work quality
+
+## Project Overview
+
+{{PROJECT_DESCRIPTION}}
+
+**Tech stack:** {{TECH_STACK}}
+**Build:** `{{BUILD_COMMAND}}`
+**Test:** `{{TEST_COMMAND}}`
+**Lint:** `{{LINT_COMMAND}}`
+
+## Key Files
+
+| File | Purpose |
+|---|---|
+| `.claude/memory.md` | Living session state — read at start, updated during work |
+| `.claude/knowledge-base.md` | Institutional memory — hard rules learned from experience |
+| `.claude/command-index.md` | All commands with triggers and tools |
+| `.claude/settings.json` | Project settings (shared with team via git) |
+| `.claude/agents/` | Specialist subagent definitions |
+| `.claude/skills/` | Domain knowledge loaded on demand |
+| `.claude/hooks/` | Deterministic enforcement scripts |
+| `.claude/logs/` | Audit trail (auto-populated by hooks) |
+
+## Retrieval Map
+
+Don't scan everything. Go straight to the right source:
+
+| You need... | Check first | Then |
+|---|---|---|
+| Current task context | `memory.md` | Daily Notes |
+| How to do a procedure | `.claude/commands/` | `.claude/skills/` |
+| What went wrong before | `knowledge-base.md` | `.claude/logs/` |
+| Project conventions | This file | Codebase itself |
+
+## Architecture Rules
+
+- Read `.claude/memory.md` before taking any action
+- Read `.claude/knowledge-base.md` at session startup
+- Keep memory.md under 100 lines — prune aggressively
+- Use skills for domain knowledge, not CLAUDE.md
+- Use hooks for enforcement, not instructions (hooks are guaranteed; instructions are advisory)
+
+## Code Conventions
+
+{{CODE_CONVENTIONS}}
+
+## Context Health
+
+Sessions have finite context. Quality degrades before the session fails.
+
+**Self-monitoring signals:**
+- After ~30 tool calls or 3+ large file reads: run `/clear`
+- If output quality degrades (repetition, missed details): run `/clear`
+- When a multi-step task completes: consider `/clear` before starting the next
+- If you see "compacting conversation": run `/clear` immediately
+
+## Memory Architecture
+
+Four layers, each with a different purpose:
+
+| Layer | File | Scope | Who writes |
+|---|---|---|---|
+| Global | `~/.claude/CLAUDE.md` | All projects | You (personal prefs) |
+| Project | `CLAUDE.md` | This project, all users | Team (committed to git) |
+| Personal | `CLAUDE.local.md` | This project, you only | You (gitignored) |
+| Dynamic | `.claude/memory.md` | Current state | Claude (updated each session) |
+
+## Command Awareness
+
+All agents can invoke commands listed in `.claude/command-index.md`.
+- **Self-execute**: If you have the tools, follow the command procedure directly
+- **Recommend**: If you lack tools, output `RECOMMEND: /command [args] — [reason]`
+
+## Maintenance
+
+- Keep this file under 150 lines (currently: ~{{LINE_COUNT}})
+- Done list cleared weekly
+- Review `.claude/logs/` during /sync
+- Auditor reviews knowledge nominations each cycle

package/templates/SETUP.md

@@ -0,0 +1,203 @@
+# Claudify Setup Guide
+
+Get from zero to a production-grade Claude Code system in under 10 minutes.
+
+---
+
+## Step 1: Install Claude Code
+
+If you haven't installed Claude Code yet:
+
+```bash
+# macOS / Linux / WSL (recommended)
+curl -fsSL https://claude.ai/install.sh | bash
+
+# Windows PowerShell
+irm https://claude.ai/install.ps1 | iex
+
+# Verify installation
+claude --version
+```
+
+**Important:** The npm install method (`npm install -g @anthropic-ai/claude-code`) is deprecated. Use the native installer above — it's faster, has zero dependencies, and auto-updates.
+
+## Step 2: Authenticate
+
+```bash
+claude
+```
+
+This opens a browser for OAuth login. You need a paid plan:
+- **Claude Pro** ($20/mo) — shared usage with claude.ai web
+- **Claude Max** ($100 or $200/mo) — 5x or 20x usage
+- **Claude Teams/Enterprise** — centralized billing
+
+**Gotcha:** If you have an `ANTHROPIC_API_KEY` environment variable set, Claude Code will use that (API billing) instead of your subscription. Unset it if you want subscription billing:
+```bash
+unset ANTHROPIC_API_KEY
+```
+
+## Step 3: Install Your Starter Pack
+
+Copy the starter pack files into your project:
+
+```bash
+# From your project root
+cp -r path/to/starter-pack/CLAUDE.md .
+cp -r path/to/starter-pack/CLAUDE.local.md .
+cp -r path/to/starter-pack/.claude .
+cp path/to/starter-pack/.mcp.json .
+```
+
+Add the gitignore entries:
+```bash
+cat path/to/starter-pack/.gitignore-additions >> .gitignore
+```
+
+## Step 4: Customize Your CLAUDE.md
+
+Open `CLAUDE.md` and replace the `{{PLACEHOLDERS}}`:
+
+1. `{{PROJECT_NAME}}` — your project name
+2. `{{PROJECT_DESCRIPTION}}` — one paragraph about what this project does
+3. `{{TECH_STACK}}` — e.g., "Next.js 15, TypeScript, Tailwind, Prisma, PostgreSQL"
+4. `{{BUILD_COMMAND}}` — e.g., `npm run build`
+5. `{{TEST_COMMAND}}` — e.g., `npm test`
+6. `{{LINT_COMMAND}}` — e.g., `npm run lint`
+7. `{{CODE_CONVENTIONS}}` — your coding standards (or delete the section)
+
+**Pro tip:** Keep CLAUDE.md under 150 lines. If you're tempted to add more, create a skill in `.claude/skills/` instead.
+
+## Step 5: Make Hook Scripts Executable
+
+```bash
+chmod +x .claude/hooks/*.sh
+```
+
+## Step 6: Configure Permissions
+
+Start Claude and set up permissions for your common commands:
+
+```bash
+claude
+# Then inside Claude:
+/permissions
+```
+
+Add your safe commands to the allow list. The starter pack pre-configures sensible defaults in `.claude/settings.json`, including:
+- Read-only tools (always allowed)
+- Git read commands (status, diff, log)
+- Build/test/lint commands
+- Dangerous commands blocked (force push, hard reset)
+
+## Step 7: Start Your First Session
+
+```bash
+claude
+```
+
+Then type `/start` to initialize. Claude will:
+1. Read your memory and knowledge base
+2. Create today's daily note
+3. Report what it knows about your project
+4. Ask what you want to work on
+
+---
+
+## What You Got
+
+### File Structure
+
+```
+your-project/
+├── CLAUDE.md                        # Project instructions (shared)
+├── CLAUDE.local.md                  # Personal overrides (gitignored)
+├── .mcp.json                        # MCP server configuration
+└── .claude/
+    ├── settings.json                # Project settings + hooks
+    ├── memory.md                    # Living session state
+    ├── knowledge-base.md            # Institutional memory
+    ├── knowledge-nominations.md     # Candidate learnings pipeline
+    ├── command-index.md             # Command registry
+    ├── commands/
+    │   ├── start.md                 # Begin a session
+    │   ├── sync.md                  # Mid-session refresh
+    │   ├── clear.md                 # Flush and resume
+    │   ├── wrap-up.md               # End a session
+    │   └── audit.md                 # Quality review
+    ├── agents/
+    │   ├── auditor.md               # Quality gate agent
+    │   └── reviewer.md              # Read-only reviewer agent
+    ├── agent-memory/
+    │   └── auditor.md               # Auditor's persistent memory
+    ├── skills/
+    │   ├── research/SKILL.md        # Research procedures
+    │   ├── deploy/SKILL.md          # Deployment procedures
+    │   ├── monitor/SKILL.md         # Health monitoring
+    │   ├── data-export/SKILL.md     # Data export procedures
+    │   └── content/SKILL.md         # Content creation
+    ├── hooks/
+    │   ├── quality-gate.sh          # Blocks incomplete writes
+    │   ├── context-monitor.sh       # Tracks context health
+    │   └── audit-logger.sh          # Records all tool calls
+    └── logs/
+        └── (auto-populated)
+```
+
+### The Four-Layer Memory System
+
+| Layer | What | Purpose |
+|---|---|---|
+| **CLAUDE.md** | Project instructions | What Claude should know about THIS project |
+| **memory.md** | Session state | What's happening RIGHT NOW |
+| **knowledge-base.md** | Hard rules | What we've LEARNED (never repeat mistakes) |
+| **knowledge-nominations.md** | Candidate rules | What we THINK we learned (auditor validates) |
+
+### The Hook System
+
+Hooks are deterministic — they run every time, unlike instructions which Claude can overlook:
+
+| Hook | When | What |
+|---|---|---|
+| **Quality gate** | Before any file write | Blocks placeholders in critical files |
+| **Context monitor** | After every response | Warns when context is getting heavy |
+| **Audit logger** | After every tool call | Records what Claude did for traceability |
+
+### The Command System
+
+Commands are reusable procedures. Type the command name and Claude follows the steps:
+
+| Command | When to use |
+|---|---|
+| `/start` | Beginning of a work session |
+| `/sync` | Mid-session to refresh context |
+| `/clear` | When context feels heavy or between tasks |
+| `/wrap-up` | End of a work session |
+| `/audit` | After finishing something important |
+
+---
+
+## Daily Workflow
+
+```
+Morning:    /start → work → /sync (if switching tasks)
+Afternoon:  work → /clear (if context heavy) → work
+Evening:    /wrap-up
+```
+
+## Tips for Getting the Most Out of Your System
+
+1. **Run `/clear` between unrelated tasks.** Context pollution is the #1 cause of quality degradation.
+2. **Keep memory.md lean.** If it's over 50 lines, you're hoarding — prune.
+3. **Let the knowledge base grow organically.** Don't pre-fill it. Let the auditor promote real learnings.
+4. **Use skills for domain knowledge.** CLAUDE.md is a routing file, not a knowledge dump.
+5. **Trust the hooks.** They catch what instructions miss.
+6. **Give Claude a way to verify.** Tests, screenshots, expected outputs — don't be the only feedback loop.
+
+---
+
+## Need Help?
+
+- Documentation: https://claudify.dev/docs
+- Support: support@claudify.dev
+- Updates ship weekly — your system gets better automatically.