prjct-cli 0.4.9 → 0.5.1

package/CHANGELOG.md

@@ -7,6 +7,164 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.5.1] - 2025-10-02
+
+### Fixed
+- **Critical: `prjct start` Command Error** - Fixed "commandInstaller.detectEditors is not a function" error
+  - **Root Cause**: v0.5.0 refactored `command-installer.js` to Claude-only architecture but didn't update calling code
+  - **Files Modified**: `core/commands.js` (lines 2227-2374)
+  - **Changes**:
+    - Refactored `start()` function to use `detectClaude()` instead of `detectEditors()`
+    - Replaced `interactiveInstall()` with `installCommands()`
+    - Removed multi-editor logic (Cursor, Windsurf references)
+    - Updated messaging to be Claude-only with download link
+    - Simplified `setup()` function with installation status checking
+    - Added `--force` flag support for reinstallation
+  - **Result**: Both `prjct start` and `prjct setup` now work correctly
+  - **Commands Installed**: Successfully installs all 21 /p:* commands to `~/.claude/commands/p/`
+
+### Fixed
+- **Corrected Claude Subscription Messaging** - Clarified honest pricing throughout
+  - Updated `website/src/pages/Changelog.tsx`:
+    - Changed "Claude Code is 100% free" → "Works with whatever Claude subscription you have (free tier or Pro)"
+    - Added clarity: "No extra costs or tokens required - just install and use with your existing Claude access"
+  - Updated `website/src/components/FAQ.tsx`:
+    - Replaced misleading "100% free" claims with accurate subscription info
+    - Emphasized "no extra setup", "no token management", "no API keys to configure"
+  - Updated `README.md`:
+    - Changed "100% free" → "works with free tier or Pro"
+    - Added clarity about zero additional costs beyond existing Claude subscription
+    - Explained Claude has free tier but isn't completely free
+    - **Editor Installation**: Removed multi-editor references (Cursor, Windsurf, Codex) - now shows Claude-only installation
+    - **Natural Language Section**: Renamed to "p. Trigger - Zero Memorization" and updated all examples to show p. prefix
+    - **Alignment**: All examples now consistently show "p. I want..." format
+    - **Data Storage**: Fixed inconsistent paths - changed `.prjct/` → `~/.prjct-cli/projects/{id}/` for accuracy
+    - **Team Workflow**: Updated to reflect global architecture with local data and shared config
+    - **FAQ Sections**: Both quick and detailed FAQs now aligned with global architecture
+  - **Why This Matters**: Being honest about costs builds trust - prjct-cli doesn't require extra payments, but Claude itself has subscription tiers
+
+- **Fixed TypeScript Errors** - Removed unused imports across website components
+  - Fixed `website/src/pages/Changelog.tsx`: Added missing `MessageSquare` import
+  - Fixed `website/src/components/Features.tsx`: Removed unused `Cpu` import
+  - Fixed `website/src/pages/Commands.tsx`: Removed unused `Cpu`, `Wind` imports
+  - Fixed `website/src/components/WindsurfExtension.tsx`: Removed unused `Zap`, `ChevronRight` imports
+  - Fixed `website/src/components/WindsurfPreview.tsx`: Removed unused `ArrowRight`, `GitBranch`, `PlayCircle`, `Layers`, `ChevronRight` imports
+  - Website now compiles without TypeScript errors
+
+### Changed
+- **Website Alignment with p. Trigger** - ALL website components now show p. trigger correctly
+  - Updated `website/src/components/Features.tsx`:
+    - Changed "p. Trigger" description to show multiple examples
+    - '"p. I\'m done" → /p:done | "p. start building auth" → /p:now | "p. ship this" → /p:ship'
+  - Updated `website/src/components/ClaudeSuperpowers.tsx`:
+    - Changed "Natural Language" → "p. Trigger - Zero Memorization"
+    - All examples now show p. prefix: "p. I'm done", "p. start building auth", "p. ship this feature"
+  - Updated `website/src/pages/Commands.tsx`:
+    - Natural language examples now include p. trigger
+    - "p. I want to start building the login page", "p. I'm done", "p. ship this feature"
+  - Updated `website/src/pages/docs/QuickStart.tsx`:
+    - Changed examples to use p. trigger
+    - "p. I'm done" when finished, "p. ship this feature" to celebrate
+
+- **Copy Simplification for Creators** - Made all text stupidly simple to understand
+  - Updated `README.md` to speak to creators and small teams, not just developers:
+    - Changed "indie hackers" → "solo creators and founders"
+    - Changed "developer momentum" → "track progress, not meetings"
+    - Simplified technical jargon:
+      - "Git validation" → "Checks your actual code changes"
+      - "MCP Integration" → "AI tools that help you code"
+      - "Dynamic AI Agents" → "Smart AI helpers"
+    - Made benefits easier to understand for non-technical creators
+  - Updated `website/src/components/InteractiveTerminal.tsx` examples:
+    - "implement user authentication" → "build login feature"
+    - "JWT token not validating" → "login not working"
+    - "API rate limiting" → "prevent spam requests"
+    - "Design notification architecture" → "Design how notifications will work"
+    - "Setup WebSocket server" → "Setup real-time messaging"
+  - Added explanatory comments to `website/src/lib/utils.ts`:
+    - Simple explanation of what `cn()` function does
+    - Real-world examples for creators
+
+### Added
+- **p. Trigger System** - Zero memorization interface for prjct context
+  - Simple prefix `p.` signals prjct context (e.g., "p. analiza esto")
+  - Works in any language: English, Spanish, German, etc.
+  - Auto-validates `.prjct/prjct.config.json` before execution
+  - Context-aware: Only works in prjct directories
+  - Provides friendly error when not in prjct project
+  - Implemented via CLAUDE.md instructions (no SDK needed)
+  - Zero API keys, zero additional costs
+
+- **Natural Language Context Validation** - Conversational responses when context is missing
+  - Automatic context validation before command execution
+  - Friendly guidance instead of technical errors
+  - Multi-language intent detection (English, Spanish, etc.)
+  - Works natively in Claude Code and Claude Desktop
+
+### Changed
+- **Natural Language System** - Enhanced with p. trigger and context validation
+  - Updated `CLAUDE.md` with complete p. trigger detection logic
+  - Added context validation examples and patterns
+  - Implemented semantic intent understanding (Claude's native LLM capability)
+  - Updated `templates/commands/done.md` with context validation docs
+  - Updated `templates/examples/natural-language-examples.md`:
+    - Added Example 11: p. Trigger (zero memorization)
+    - Added Example 12: p. Trigger multi-language
+    - Added Example 13: p. Trigger without project
+    - Added Example 14: All three ways work (p. | /p: | natural)
+    - Added examples for missing context scenarios (Example 2b, 3b)
+    - Updated implementation notes (no SDK approach)
+  - Created `docs/p-trigger.md` - Complete p. trigger documentation
+
+- **Website Alignment** - Updated all components to show p. trigger as primary interface
+  - `website/src/components/Hero.tsx`:
+    - Updated p. Trigger feature description
+    - Changed badge to "Built for Claude Code - Claude Desktop"
+    - Restored Windsurf Extension CTA with scroll-to-section functionality
+    - Added Sparkles and ArrowRight animated icons
+  - `website/src/components/Features.tsx`:
+    - Updated p. Trigger feature to show natural language first
+  - `website/src/components/CommandGuide.tsx`:
+    - Added p. trigger examples to all 6 scenarios (preferred option)
+    - Updated Quick Decision Matrix to show p. trigger format
+    - All examples now follow: `"p. [natural language]"` → `/p:command` pattern
+    - Demonstrates zero memorization approach throughout
+
+- **Windsurf Extension Preview Section** - Added complete Windsurf extension roadmap to homepage
+  - `website/src/components/WindsurfExtension.tsx`:
+    - Full section with hero, features grid, timeline, and early access form
+    - Development timeline (Oct 2025 - Feb 2026)
+    - 4 key features: Real-time Metrics, Focus Mode, Velocity Tracking, Smart Notifications
+    - Progress bar showing 10% completion (validation phase)
+    - Scroll target for Hero CTA (`.windsurf-extension-section`)
+  - `website/src/components/WindsurfPreview.tsx`:
+    - Interactive mockup of Windsurf extension UI
+    - Kanban-style roadmap with drag & drop
+    - AI Task Generator and PRJCT Control Center
+    - Project Rules management interface
+  - `website/src/components/EarlyAccessForm.tsx`:
+    - GitHub issue-based waitlist signup
+    - Animated with framer-motion
+    - Purple/blue gradient theme
+  - `website/src/pages/Home.tsx`:
+    - Added WindsurfExtension between Features and ClaudeSuperpowers
+    - Maintains page flow and user journey
+
+### Fixed
+- **🐛 Natural Language Context Bug** - Fixed error messages when context is missing
+  - Before: "⏺ No hay tarea activa en este momento..." (technical error)
+  - After: "✨ You're not working on anything right now! Want to start something?" (conversational)
+  - Affects ALL commands: /p:done, /p:ship, /p:stuck, /p:idea, /p:now
+  - Claude automatically validates and provides friendly guidance
+
+### Technical Implementation
+- **Zero External Dependencies** - All functionality via CLAUDE.md
+  - No SDK required (initially considered but removed)
+  - Uses Claude Code's native session (zero API keys)
+  - Simple file reads for context validation
+  - Works identically in Claude Code and Claude Desktop
+  - No additional costs for users
+
 ### Coming Soon
 - **Windows Compatibility** - Native Windows support
   - PowerShell and CMD command execution
@@ -15,6 +173,180 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   - Cross-platform file operations
   - Windows Terminal integration
 
+## [0.5.0] - 2025-10-02
+
+### 🚨 BREAKING CHANGES
+
+- **100% Claude-Focused Architecture** - Removed support for all non-Claude environments
+  - ❌ **Removed**: OpenAI Codex support
+  - ❌ **Removed**: Cursor AI support
+  - ❌ **Removed**: Windsurf/Codeium support
+  - ✅ **Supported**: Claude Code (primary)
+  - ✅ **Supported**: Claude Desktop (secondary)
+
+- **Simplified Installation** - Now installs only to `~/.claude/`
+  - Single editor detection (Claude only)
+  - Removed multi-editor selection workflow
+  - Removed editor-specific adapters and formatters
+
+- **Deleted Files**:
+  - `core/agents/codex-agent.js` - OpenAI Codex adapter
+  - `core/agents/terminal-agent.js` - Terminal fallback
+  - `AGENTS.md` - Legacy Codex configuration
+  - `templates/workflows/` - Windsurf workflow templates
+
+- **Rewritten Modules** (breaking API changes):
+  - `command-installer.js` - Now Claude-only (208 lines, was 800+)
+  - `editors-config.js` - Simplified to single editor (160 lines, was 205)
+  - `agent-detector.js` - Claude detection only (183 lines, was 250)
+
+### 🚀 Why This Change? (Built for Claude - Ship Fast, No BS)
+
+**prjct-cli is now a developer momentum tool, NOT a project management tool.**
+
+**The Philosophy:**
+- Ship fast, stay focused, no BS
+- For indie hackers and small teams (2-5 devs)
+- No Jira, no ceremonies, no bureaucracy
+- Just `/p:now` → work → `/p:done` → `/p:ship` → celebrate
+
+**By focusing 100% on Claude, we unlocked superpowers:**
+
+1. **🤖 Dynamic AI Agents**
+   - Auto-generated specialists (PM, Frontend, Backend, UX, QA, Scribe, Security, DevOps, Mobile, Data)
+   - Context-aware activation based on your project stack
+   - Impossible to build with multi-platform support
+
+2. **🔗 Native MCP Integration**
+   - Context7 - Automatic library documentation
+   - Sequential - Deep reasoning for complex problems
+   - Magic - UI component generation from patterns
+   - Playwright - Browser automation and E2E testing
+   - MCP is Claude-native, can't replicate elsewhere
+
+3. **✅ Git Validation**
+   - Last commit as source of truth
+   - Validates completed work against actual changes
+   - Prevents empty task completions
+   - Requires tight Claude integration
+
+4. **💬 Natural Language**
+   - Talk naturally in any language, no command memorization
+   - "I'm done" → `/p:done`, "ship this" → `/p:ship`
+   - Leverages Claude's language understanding
+
+5. **⚡ Technical Benefits**
+   - **50-60% less code** (800+ lines → 228 lines in command-installer alone)
+   - Faster features and bug fixes
+   - Proper testing of everything we support
+   - Honest compatibility (only claim what works)
+
+**Migration Path:**
+- **Using Claude?** → `npm update -g prjct-cli` (you're all set!)
+- **Using Cursor/Windsurf/Codex?** → Switch to Claude Code (free) or stay on v0.4.10
+- **See [MIGRATION.md](MIGRATION.md) for complete upgrade guide**
+
+**This isn't a limitation - it's a strategic decision that makes prjct-cli better for developers who ship fast.**
+
+### Added
+- **Dynamic AI Agent System** - Automatic generation of specialized AI agents based on project stack
+  - **GitIntegration Module** (`core/git-integration.js`)
+    - Repository validation and state tracking
+    - Last commit as source of truth for validation
+    - Working directory status monitoring
+    - User claim validation against commit history
+    - Comprehensive git statistics and reporting
+
+  - **AgentGenerator Module** (`core/agent-generator.js`)
+    - Dynamic agent creation based on project analysis
+    - Project-specific context injection into agent templates
+    - Intelligent conditional agent detection
+    - Agent update and synchronization capabilities
+    - Support for both base and conditional agents
+
+  - **Base AI Agents** (Always Generated)
+    - PM (Project Manager) - Task coordination and breakdown
+    - FE (Frontend Engineer) - UI/UX, React, components, state management
+    - BE (Backend Engineer) - APIs, database, SOLID principles
+    - UX (UX Designer) - User experience, accessibility, design systems
+    - QA (QA Engineer) - Testing strategy and automation
+    - Scribe (Documentation) - Technical writing and changelog management
+
+  - **Conditional AI Agents** (Generated Based on Project Stack)
+    - DevOps - Docker, CI/CD, infrastructure (when Docker/CI/CD detected)
+    - Security - OWASP, threat modeling (when web app or auth detected)
+    - Mobile - React Native, Flutter (when mobile frameworks detected)
+    - Data - ML, data pipelines (when ML libraries detected)
+
+  - **Scribe Agent Workflows**
+    - Automatic documentation on `/p:done` completion
+    - Feature documentation on `/p:ship` deployment
+    - Git-based change detection for documentation scope
+    - User confirmation workflow before saving documentation
+    - Task docs saved to `analysis/task-docs/`
+    - Feature docs saved to `analysis/feature-docs/`
+
+- **Enhanced `/p:analyze` Command**
+  - Git repository integration and validation
+  - Last commit information and statistics
+  - Working directory status tracking
+  - Validation baseline for user claims
+  - Conditional agent detection based on stack
+  - Comprehensive project analysis report with git stats
+  - Agent recommendation system
+
+- **Enhanced `/p:init` Command**
+  - Automatic agent generation after project analysis
+  - Git-integrated analysis workflow
+  - Project-specific agent context injection
+  - Smart agent selection based on detected stack
+  - Agents stored in `~/.claude/agents/` directory
+
+- **New `/p:sync` Command**
+  - Re-analyze project for stack changes
+  - Update existing agents with new project context
+  - Add newly required agents based on stack evolution
+  - Remove obsolete agents with user confirmation
+  - Comprehensive sync reporting and logging
+
+### Changed
+- **Agent Architecture** - 100% Claude-focused (Claude Code + Claude Desktop)
+  - Agents stored in `~/.claude/agents/` directory
+  - Markdown format with YAML frontmatter (Claude native)
+  - Project-specific naming: `p_agent_[type].md`
+  - Dynamic context injection from project analysis
+  - Limited expertise per agent (focused specialization)
+
+### Technical Details
+- **10 Agent Templates** in `templates/agents/`
+  - 6 base templates (always generated)
+  - 4 conditional templates (stack-dependent)
+  - Each template includes project context placeholders
+  - Specialized tools and color coding per agent type
+  - Defined expertise boundaries and trigger keywords
+
+- **Git Integration Features**
+  - Commit validation and claim verification
+  - Working directory change tracking
+  - Unverified work identification
+  - Comprehensive git statistics
+  - Multi-contributor support
+
+## [0.4.10] - 2025-10-02
+
+### Changed
+- **Agent Workflow Initialization** - `prjct init` now properly initializes AI agent workflows
+  - When existing codebase detected, prompts user to run `/p:analyze`
+  - Enables AI agents to follow complete analysis workflow
+  - Provides proper workflow context for comprehensive project analysis
+  - Replaced silent programmatic analysis with workflow-driven approach
+
+### Fixed
+- **Analysis Workflow** - AI agents now receive proper instructions during project initialization
+  - Removed silent `analyze()` call that bypassed agent workflows
+  - Agent workflows now activate when user runs `/p:analyze` command
+  - Ensures agents follow complete analysis instructions from workflow templates
+
 ## [0.4.9] - 2025-10-02
 
 ### Added

package/CLAUDE.md

@@ -46,6 +46,35 @@ The system operates as an AI Assistant Enhancement Framework using:
 
 **prjct-cli includes a conversational interface** - users can talk naturally instead of memorizing commands!
 
+### p. Trigger Detection
+
+**NEW: Simple prefix trigger** - Users can start messages with `p.` to signal prjct context:
+
+**How it works:**
+1. User starts message with `p.` → You check if `.prjct/prjct.config.json` exists
+2. If exists → Detect intent from rest of message → Execute appropriate `/p:*` command
+3. If not exists → Respond: "No prjct project here. Run /p:init first."
+
+**Examples:**
+```
+"p. analiza todo este documento" → Detect "analysis" intent → /p:analyze
+"p. estoy listo para shipear" → Detect "ship" intent → /p:ship
+"p. muéstrame progreso semanal" → Detect "progress" intent → /p:progress week
+"p. empiezo con auth" → Detect "start task" intent → /p:now "auth"
+"p. terminé" → Detect "done" intent → /p:done
+```
+
+**Why this is useful:**
+- ✅ **Zero memorization** - No need to remember `/p:*` commands
+- ✅ **Multi-project** - Only works in prjct directories
+- ✅ **Natural** - "p. [what you want]" is intuitive
+- ✅ **Any language** - Works in English, Spanish, etc.
+
+**Detection priority:**
+1. Check for `p.` prefix first
+2. Then check for `/p:*` slash command
+3. Finally check for natural language without prefix
+
 ### Semantic Intent Detection
 
 As an LLM, you understand context and intent naturally. When users describe what they want to do, **map their intent to the appropriate command** based on semantic understanding, not pattern matching.
@@ -101,11 +130,11 @@ What's next?
 Or use: /p:now | /p:ship | /p:idea
 ```
 
-## Implementation for Claude Code
+## Implementation for Claude Code & Desktop
 
 ### How to Handle Natural Language
 
-When you receive a message from the user:
+When you receive a message from the user, follow these steps to provide the natural language experience:
 
 **Step 1: Check if it's a direct command**
 ```javascript
@@ -142,7 +171,46 @@ if (intent.command) {
 }
 ```
 
-**Step 3: Provide conversational response**
+**Step 3: Validate context before execution**
+```javascript
+// Before executing commands, check prerequisites:
+
+// For /p:done:
+if (command === 'done') {
+  const nowContent = await Read('core/now.md')
+  if (!nowContent || nowContent.trim() === '') {
+    return conversationalResponse(`
+✨ You're not working on anything right now!
+
+Want to start something?
+• Tell me what you want to build
+• Say "show me what's next"
+• Or use: /p:now | /p:next
+
+Let's ship something!`)
+  }
+}
+
+// For /p:ship:
+if (command === 'ship') {
+  const nowContent = await Read('core/now.md')
+  const shippedContent = await Read('progress/shipped.md')
+  if ((!nowContent || nowContent.trim() === '') &&
+      (!shippedContent || shippedContent.trim() === '')) {
+    return conversationalResponse(`
+🚀 Nothing to ship yet!
+
+First, let's build something:
+• Tell me what feature you want to create
+• Say "start building [feature]"
+• Or use: /p:now "feature name"
+
+Then we'll celebrate when you ship it!`)
+  }
+}
+```
+
+**Step 4: Provide conversational response**
 ```javascript
 // Always suggest natural next steps
 ✅ [What you did]
@@ -178,6 +246,43 @@ When you're done:
 Need help? Say "I'm stuck" or use /p:stuck
 ```
 
+### Context Validation Examples
+
+**Missing Active Task:**
+
+User: "done"
+You check: `core/now.md` is empty
+Your response:
+```
+✨ You're not working on anything right now!
+
+Want to start something?
+• Tell me what you want to build
+• Say "show me what's next"
+• Or use: /p:now | /p:next
+
+Let's ship something!
+```
+
+**Valid Context:**
+
+User: "I'm done"
+You check: `core/now.md` has content ✅
+Your response:
+```
+💬 I understood: "task completion"
+⚡ Executing: /p:done
+
+✅ Task complete: building authentication (2h 15m)
+
+What's next?
+• "start next task" → Begin working
+• "ship this feature" → Track & celebrate
+• "add new idea" → Brainstorm
+
+Or use: /p:now | /p:ship | /p:idea
+```
+
 ### Key Principles
 
 1. **Trust your understanding** - You're an LLM, not a regex parser
@@ -185,6 +290,7 @@ Need help? Say "I'm stuck" or use /p:stuck
 3. **Variations are fine** - Users can express intent in infinite ways
 4. **Show transparency** - Always say what you understood and what you'll execute
 5. **Be conversational** - Guide users naturally toward next actions
+6. **Validate context** - Check prerequisites before executing commands
 
 ## Command System for Claude Code