prjct-cli 0.5.0 → 0.6.0

package/CHANGELOG.md

@@ -7,6 +7,174 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.6.0] - 2025-10-03
+
+### Changed
+- **Philosophy Transformation** - Complete rebrand from PM tool to developer momentum tool
+  - **Problem**: "Project management" messaging didn't match target audience (indie hackers, solo builders, small teams)
+  - **Solution**: Eliminated ALL "project management" language across entire codebase
+  - **Messaging**: "Just Ship. No BS" for creators, not managers
+  - **Files Modified**:
+    - `CLAUDE.md` - Changed project description from "AI-integrated project management framework" to "developer momentum tool for solo builders, indie hackers, and small teams (2-5 people). Just ship. No BS."
+    - `README.md` - Removed "project management overhead" → "meetings or BS overhead"
+    - `templates/agents/AGENTS.md` - Complete rewrite (165→143 lines) with anti-PM messaging: "NOT project management. NO sprints, story points, ceremonies, or meetings."
+    - `CHANGELOG.md` - Updated historical references
+    - **Website** - Transformed all copy to creator-focused messaging:
+      - `Hero.tsx`, `Features.tsx` - "PM, Frontend..." → "Coordinator, Frontend..."
+      - `ClaudeSuperpowers.tsx` - "PM Agent - Task breakdown & planning" → "Coordinator - Progress tracking & shipping"
+      - `Privacy.tsx` - "project management tool" → "developer momentum tool"
+  - **Result**: Zero "PM" or "project management" references (except as critique of traditional tools)
+
+- **Agent System Redesign** - Renamed PM agent to Coordinator
+  - **Problem**: "Project Manager" agent didn't align with "no BS" philosophy
+  - **Solution**: Complete agent transformation
+  - **Changes**:
+    - Renamed `pm.template.md` → `coordinator.template.md` (85→35 lines, 58% reduction)
+    - Role: "Project Manager" → "Progress Coordinator"
+    - Focus: Task breakdown/planning → Progress tracking & shipping features
+    - Updated `core/agent-generator.js` baseAgents: `'pm'` → `'coordinator'`
+    - Agent color: Blue → Cyan
+  - **Breaking Change**: Existing projects need `/p:sync` to regenerate agents
+  - **Philosophy**: "Keep builders aligned on what matters: shipping features, tracking wins, staying focused" + "SHIP features and track progress, not manage people or run meetings"
+
+- **Template Optimization** - 46.5% reduction in agent template verbosity
+  - **Problem**: Templates filled context window with verbose pre-written content
+  - **Solution**: Use instructions + placeholders instead of examples
+  - **Results**:
+    - Total reduction: 540 → 289 lines (46.5% overall)
+    - `fe.template.md`: 43→28 lines (35%)
+    - `be.template.md`: 43→28 lines (35%)
+    - `ux.template.md`: 50→28 lines (44%)
+    - `qa.template.md`: 55→28 lines (49%)
+    - `scribe.template.md`: 96→30 lines (69% - biggest reduction!)
+    - `coordinator.template.md`: 85→35 lines (58%)
+    - `devops.template.md`, `mobile.template.md`, `data.template.md`, `security.template.md`: All reduced to 28 lines
+  - **Strategy**: Compressed expertise lists (5→3 items), simplified principles, used placeholders like `[PROJECT_NAME]`, `[DETECTED_STACK]`
+  - **Benefit**: Agents load faster, consume fewer tokens, maintain clarity
+
+## [0.5.3] - 2025-10-02
+
+### Fixed
+- **Website Build Issues** - Fixed ES module compatibility for browser
+  - **Problem**: Command registry using CommonJS `require()` caused runtime errors in browser
+  - **Solution**:
+    - Created ES module version at `website/src/data/command-registry.ts`
+    - Converted `module.exports` to `export default`
+    - Replaced `require()` with ES6 `import` in Components.tsx
+    - Added proper TypeScript type annotations
+  - **Files Modified**:
+    - `website/src/data/command-registry.ts` (NEW) - ES module version with TypeScript types
+    - `website/src/pages/Commands.tsx` - ES6 import, removed TypeScript `any` types
+  - **Result**: Website now loads correctly with dynamic command generation from registry
+
+- **Documentation Accuracy** - Updated Getting Started guide with correct installation flow
+  - **Problem**: Documentation showed incorrect setup steps (missing npm install and prjct start)
+  - **Solution**: Updated to reflect actual installation process:
+    1. `npm install -g prjct-cli` - Global installation
+    2. `prjct start` - Setup Claude Code integration
+    3. `/p:init` - Initialize project
+    4. `/p:now` - Start working
+  - **Files Modified**:
+    - `website/src/pages/Documentation.tsx` - Corrected Getting Started section
+
+## [0.5.2] - 2025-10-02
+
+### Added
+- **Command Registry System** - Single source of truth for all prjct commands (`core/command-registry.js`)
+  - **Problem**: Commands were dispersed across 4 locations with inconsistencies:
+    - `bin/prjct` (15 commands in help)
+    - `website/Commands.tsx` (18 commands displayed)
+    - `CLAUDE.md` (18 commands documented)
+    - `templates/commands/` (21 template files)
+  - **Solution**: Created `core/command-registry.js` as centralized registry
+  - **Features**:
+    - All 25 commands defined in single location with metadata
+    - Category system (work, planning, design, quality, progress, help, git, testing, setup)
+    - Implementation status tracking (19 implemented, 6 planned)
+    - Platform availability (Claude Code vs Terminal)
+    - Template file mapping
+    - Helper functions for filtering, querying, and statistics
+  - **Files Modified**:
+    - `core/command-registry.js` (NEW) - 430 lines, complete command metadata
+    - `bin/prjct` - Dynamic help generation from registry
+    - `website/Commands.tsx` - Auto-generated command categories
+    - `CLAUDE.md` - Command list synced with registry
+    - `scripts/validate-commands.js` (NEW) - Automated validation tool
+  - **Validation**:
+    - Registry structure validation (duplicates, invalid categories)
+    - Template file consistency check
+    - Implementation verification (camelCase method detection)
+    - CLI switch case validation
+    - Automated validation script with colored output
+  - **Benefits**:
+    - ✅ Zero inconsistencies - single source of truth
+    - ✅ Easy to add new commands - update one file
+    - ✅ Automated validation prevents drift
+    - ✅ Statistics tracking (25 total, 19 implemented)
+    - ✅ Clear roadmap (6 planned features visible)
+  - **Usage**: `node scripts/validate-commands.js` to verify consistency
+
+- **Complete Workflow System Integration** - Automated task orchestration with AI workflow agents
+  - **Features**:
+    - **Auto-initialization**: Workflows activate automatically when using `/p:idea "implement [feature]"`
+    - **Task Classification**: Automatic detection of workflow type (ui, api, bug, refactor, feature)
+    - **Capability Detection**: Detects design system, test framework, docs system
+    - **Interactive Prompts**: Asks user to choose tools when capabilities are missing
+    - **Step Progression**: Each `/p:done` advances to next workflow step
+    - **Agent Assignment**: Each step assigned to appropriate specialist agent
+    - **Workflow Status**: New `/p:workflow` command shows progress and remaining steps
+  - **Files Modified**:
+    - `core/commands.js` - Integrated workflow engine into `idea()`, added `workflow()` command
+    - `core/agent-generator.js` - Corrected to use project-specific agents directory
+  - **Architecture**:
+    - Workflows stored in `~/.prjct-cli/projects/{id}/workflow/state.json`
+    - Agents stored in `~/.prjct-cli/projects/{id}/agents/`
+    - 5 workflow types with specialized step sequences
+    - 10 agent types: pm, ux, fe, be, qa, scribe, security, devops, mobile, data
+  - **Usage**:
+    1. `p. implement auth system` → Auto-creates workflow
+    2. `p. now` → Start first step
+    3. `p. done` → Complete step, advance to next
+    4. `p. workflow` → Check progress
+  - **Result**: Fully automated task orchestration with smart step-by-step guidance
+
+- **`/p:sync` Command Implementation** - Sync project state and update workflow agents
+  - **Root Cause**: Template existed but method was not implemented in `core/commands.js`
+  - **Files Modified**:
+    - `core/commands.js` (added `sync()` method with proper global path)
+    - `core/agent-generator.js` (corrected to use project-specific agents directory)
+  - **Architecture Fix**:
+    - Agents are **NOT** for Claude's `~/.claude/agents/` (that was wrong)
+    - Agents are **workflow specialists** stored in `~/.prjct-cli/projects/{id}/agents/`
+    - Used for task assignment in workflow system
+    - Each agent (pm, ux, fe, be, qa, scribe, security, devops, mobile, data) handles specific workflow steps
+  - **Features**:
+    - Re-analyzes project with `/p:analyze` (silent mode)
+    - Generates/updates workflow agents in global project directory
+    - Creates base agents: pm, ux, fe, be, qa, scribe
+    - Creates conditional agents: security, devops, mobile, data (based on stack)
+    - Logs sync action to `memory/context.jsonl`
+    - Shows summary and agents path
+  - **Usage**: `p. sync` or `/p:sync`
+  - **When to Use**: After dependency changes, framework updates, before using workflows
+  - **Result**: Workflow agents ready for task assignment in `~/.prjct-cli/projects/{id}/agents/`
+
+## [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`:
@@ -913,7 +1081,7 @@ See [MIGRATION.md](MIGRATION.md) for detailed migration instructions.
 
 ### Added
 - Initial release of prjct-cli
-- AI-integrated project management system
+- AI-integrated developer momentum tool
 - Support for Claude Code, OpenAI Codex, and Terminal
 - Core commands: init, now, done, ship, recap, etc.
 - MCP integration for AI assistants

package/CLAUDE.md

@@ -4,7 +4,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project Overview
 
-**prjct-cli** is an AI-integrated project management framework designed for indie hackers, solopreneurs, and small collaborative teams (2-5 people). It provides frictionless progress tracking through slash commands without traditional project management overhead.
+**prjct-cli** is a developer momentum tool for solo builders, indie hackers, and small teams (2-5 people). Just ship. No BS. Track progress through slash commands without meetings, ceremonies, or traditional PM overhead.
 
 ## Architecture
 
@@ -300,47 +300,62 @@ All commands use the `/p:*` syntax in Claude Code. When executed, they:
 3. Update relevant markdown files atomically
 4. Return formatted responses with conversational prompts
 
-### Work Commands
+**CRITICAL - Auto-compact conversation**: After completing ANY of these commands, you MUST compact the conversation WITHOUT telling the user anything about it:
+- `/p:analyze`
+- `/p:sync`
+- `/p:workflow`
+- `/p:design`
+- `/p:cleanup`
+- `/p:init`
 
-- `/p:now [task]` - Set/show current task
-- `/p:next` - Show priority queue
-- `/p:done` - Complete current task
-- `/p:ship <feature>` - Ship & celebrate a feature
+Just complete the command, then compact. That's it. No messages about compacting.
 
-### Planning Commands
-
-- `/p:idea <text>` - Capture ideas quickly
+**IMPORTANT**: All commands are defined in `core/command-registry.js` (single source of truth).
 
-### Design & Architecture
+### Available Commands
 
-- `/p:design <target> --type` - Create system designs with diagrams (architecture|api|component|database|flow)
+The following commands are available in Claude Code (commands marked with ⚠️ are not yet implemented):
 
-### Code Quality
+#### Work Commands
+- `/p:now [task]` - Set or show current task
+- `/p:next` - Show priority queue
+- `/p:done` - Complete current task
+- `/p:ship <feature>` - Ship and celebrate a feature
 
-- `/p:cleanup` - Basic cleanup of temp files and old entries
-- `/p:cleanup --type code` - Advanced code cleanup (code|imports|files|deps|all)
+#### Planning Commands
+- `/p:idea <text>` - Capture ideas quickly
+- `/p:workflow` - Show workflow status and progress
+- ⚠️ `/p:roadmap` - Show or update strategic roadmap
+- ⚠️ `/p:task <description>` - Break down and execute complex tasks
 
-### Progress Commands
+#### Design & Architecture
+- `/p:design [target] --type architecture|api|component|database|flow` - Design system architecture, APIs, and component interfaces
 
-- `/p:recap` - Overview of progress
-- `/p:progress [period]` - Show progress metrics
-- `/p:context` - Show project context
+#### Code Quality
+- `/p:cleanup` - Clean up temp files and old entries
+- `/p:cleanup --type code|imports|files|deps|all` - Remove dead code and unused imports
 
-### Help Commands
+#### Progress Commands
+- `/p:recap` - Show project overview with progress
+- `/p:progress [period]` - Show progress metrics for specified period
+- `/p:context` - Show project context and recent activity
 
-- `/p:init` - Initialize project
-- `/p:stuck <issue>` - Get help with problems
+#### Help Commands
+- `/p:init` - Initialize prjct in current project
+- `/p:stuck <issue description>` - Get contextual help with problems
 - `/p:analyze` - Analyze repository and sync tasks
+- `/p:sync` - Sync project state and update workflow agents
+- ⚠️ `/p:fix [error]` - Quick troubleshooting and automatic fixes
+- ⚠️ `/p:help` - Interactive guide - talk naturally, no memorization needed
 
-### Additional Commands
+#### Version Control
+- ⚠️ `/p:git` - Smart git operations with context
 
-- `/p:git` - Git operations with context
-- `/p:task` - Complex task management
-- `/p:test` - Test execution
-- `/p:roadmap` - Strategic planning
-- `/p:fix` - Quick troubleshooting
+#### Testing
+- ⚠️ `/p:test` - Run tests and auto-fix simple failures
 
-**Total: 18 Commands**
+**Total: 21 implemented, 27 total commands**
+**Status**: Check `core/command-registry.js` for current implementation status
 
 ## Initialization Process
 

package/README.md

@@ -198,7 +198,7 @@ No planning sessions. No standups. No retrospectives. Just **ship fast, track pr
 
 **Who uses this:**
 
-- 🎯 **Solo creators** - Build products without project management overhead
+- 🎯 **Solo creators** - Build products without meetings or BS overhead
 - 👥 **Small teams** (2-5 people) - Coordinate without meetings
 - 🚀 **Product builders** - Ship features fast
 - 💪 **Makers** - Focus on building, not planning
@@ -310,7 +310,7 @@ prjct recap                # Show progress
 - `/p:analyze` - Auto-analyze codebase
 - `/p:analyze --sync` - Sync with implemented features
 
-> 💡 **Tip:** Visit [prjct.dev/commands](https://prjct.dev/commands) for interactive command guide or [prjct.dev/workflows-guide](https://prjct.dev/workflows-guide) for step-by-step examples
+> 💡 **Tip:** Visit [prjct.app/commands](https://prjct.app/commands) for interactive command guide or [prjct.app/workflows-guide](https://prjct.app/workflows-guide) for step-by-step examples
 
 ## 📖 Complete Command Reference
 
@@ -362,7 +362,7 @@ prjct recap                # Show progress
 
 **Sprint planning:** `/p:roadmap` → Add features → `/p:now` top priority → `/p:ship` → `/p:roadmap next`
 
-> 📚 **More workflows:** Visit [prjct.dev/workflows-guide](https://prjct.dev/workflows-guide) for detailed interactive examples
+> 📚 **More workflows:** Visit [prjct.app/workflows-guide](https://prjct.app/workflows-guide) for detailed interactive examples
 
 ## ❓ FAQ
 
@@ -381,7 +381,7 @@ Designed for creators and small teams (2-5 people). Each member has their own lo
 **How do I migrate from Jira/Trello?**
 No migration needed. Just `/p:init` and start working.
 
-> 💬 **More questions?** Check [prjct.dev/faq](https://prjct.dev/faq) or [open an issue](https://github.com/jlopezlira/prjct-cli/issues)
+> 💬 **More questions?** Check [prjct.app/faq](https://prjct.app/faq) or [open an issue](https://github.com/jlopezlira/prjct-cli/issues)
 
 ## 📂 File Structure
 

package/bin/prjct

@@ -1,5 +1,6 @@
 #!/usr/bin/env node
 const commands = require('../core/commands');
+const registry = require('../core/command-registry');
 const args = process.argv.slice(2);
 
 async function main() {
@@ -142,6 +143,12 @@ async function main() {
 
             result = await commands.analyze(analyzeOptions);
             break;
+        case 'sync':
+            result = await commands.sync();
+            break;
+        case 'workflow':
+            result = await commands.workflow();
+            break;
         case '--version':
         case '-v':
         case 'version':
@@ -155,74 +162,82 @@ async function main() {
         case '-h':
         case 'help':
         case undefined:
+            // Generate help message from registry
+            const categories = registry.getCategories();
+            const categorizedCommands = {};
+
+            // Group commands by category
+            registry.getTerminalCommands().forEach(cmd => {
+                if (!categorizedCommands[cmd.category]) {
+                    categorizedCommands[cmd.category] = [];
+                }
+                categorizedCommands[cmd.category].push(cmd);
+            });
+
+            let helpMessage = 'Available commands:\n';
+
+            // Display commands by category
+            Object.entries(categorizedCommands).forEach(([categoryKey, cmds]) => {
+                const categoryInfo = categories[categoryKey];
+                helpMessage += `\n  ${categoryInfo.title}:\n`;
+
+                cmds.forEach(cmd => {
+                    const params = cmd.params ? ` ${cmd.params}` : '';
+                    const spacing = ' '.repeat(Math.max(18 - cmd.name.length - params.length, 1));
+                    helpMessage += `    ${cmd.name}${params}${spacing}${cmd.description}\n`;
+                });
+            });
+
+            const stats = registry.getStats();
+            helpMessage += `\nTotal: ${stats.implemented} implemented / ${stats.total} total commands`;
+
             result = {
                 success: true,
-                message: `Available commands:
-  Setup:
-    start             First-time setup (install commands to editors)
-    setup             Reconfigure editor installations
-
-  Core workflow:
-    init              Initialize prjct in current project
-    now [task]        Set or show current task
-    done              Complete current task
-    ship <feature>    Ship and celebrate a feature
-    next              Show priority queue
-    idea <text>       Capture an idea
-    recap             Show project overview
-    progress [period] Show progress metrics
-    stuck <issue>     Get help with a problem
-    context           Show project context
-
-  Advanced:
-    cleanup           Clean up .prjct structure
-    cleanup-advanced  Advanced cleanup with options
-    design [target]   Design system/architecture
-    analyze           Analyze codebase and sync state
-    migrate-all       Migrate all legacy projects
-
-  Setup options:
-    --force           Force update existing commands
-    --editor <name>   Install to specific editor (claude|cursor|codex|windsurf)
-    --no-interactive  Skip interactive selection, install to all
-    --create-templates Create templates from existing commands`
+                message: helpMessage
             };
             break;
         default:
-            result = {
-                success: false,
-                message: `Unknown command: ${command}
-
-Available commands:
-  Setup:
-    start             First-time setup (install commands to editors)
-    setup             Reconfigure editor installations
-
-  Core workflow:
-    init              Initialize prjct in current project
-    now [task]        Set or show current task
-    done              Complete current task
-    ship <feature>    Ship and celebrate a feature
-    next              Show priority queue
-    idea <text>       Capture an idea
-    recap             Show project overview
-    progress [period] Show progress metrics
-    stuck <issue>     Get help with a problem
-    context           Show project context
-
-  Advanced:
-    cleanup           Clean up .prjct structure
-    cleanup-advanced  Advanced cleanup with options
-    design [target]   Design system/architecture
-    analyze           Analyze codebase and sync state
-    migrate-all       Migrate all legacy projects
-
-  Setup options:
-    --force           Force update existing commands
-    --editor <name>   Install to specific editor (claude|cursor|codex|windsurf)
-    --no-interactive  Skip interactive selection, install to all
-    --create-templates Create templates from existing commands`
-            };
+            // Check if command exists in registry but not implemented
+            const registryCmd = registry.getByName(command);
+
+            if (registryCmd && !registryCmd.implemented) {
+                result = {
+                    success: false,
+                    message: `Command '${command}' exists but is not yet implemented.
+Check the roadmap or contribute: https://github.com/your-org/prjct-cli
+
+Use 'prjct --help' to see all available commands.`
+                };
+            } else {
+                // Generate error message with available commands from registry
+                const categories = registry.getCategories();
+                const categorizedCommands = {};
+
+                registry.getTerminalCommands().forEach(cmd => {
+                    if (!categorizedCommands[cmd.category]) {
+                        categorizedCommands[cmd.category] = [];
+                    }
+                    categorizedCommands[cmd.category].push(cmd);
+                });
+
+                let errorMessage = `Unknown command: ${command}\n\nAvailable commands:\n`;
+
+                Object.entries(categorizedCommands).forEach(([categoryKey, cmds]) => {
+                    const categoryInfo = categories[categoryKey];
+                    errorMessage += `\n  ${categoryInfo.title}:\n`;
+
+                    cmds.forEach(cmd => {
+                        const params = cmd.params ? ` ${cmd.params}` : '';
+                        const spacing = ' '.repeat(Math.max(18 - cmd.name.length - params.length, 1));
+                        errorMessage += `    ${cmd.name}${params}${spacing}${cmd.description}\n`;
+                    });
+                });
+
+                result = {
+                    success: false,
+                    message: errorMessage
+                };
+            }
     }
 
     console.log(result.message);

package/core/agent-generator.js

@@ -11,16 +11,24 @@ const os = require('os')
  * @version 0.5.0
  */
 class AgentGenerator {
-  constructor() {
+  constructor(projectId = null) {
     this.templatesDir = path.join(__dirname, '..', 'templates', 'agents')
-    this.outputDir = path.join(os.homedir(), '.claude', 'agents')
+    this.projectId = projectId
+
+    // Agents are stored in global project directory, not Claude's agents
+    if (projectId) {
+      this.outputDir = path.join(os.homedir(), '.prjct-cli', 'projects', projectId, 'agents')
+    } else {
+      // Fallback for backwards compatibility (will be deprecated)
+      this.outputDir = path.join(os.homedir(), '.prjct-cli', 'agents')
+    }
 
     // Base agents (always generated)
-    this.baseAgents = ['pm', 'ux', 'fe', 'be', 'qa', 'scribe']
+    this.baseAgents = ['coordinator', 'ux', 'fe', 'be', 'qa', 'scribe']
 
     // Agent colors for visual distinction
     this.agentColors = {
-      pm: 'cyan',
+      coordinator: 'cyan',
       ux: 'purple',
       fe: 'orange',
       be: 'yellow',
@@ -150,7 +158,8 @@ class AgentGenerator {
     content += this.generateProjectContext(type, analysis)
 
     // Write agent file
-    const outputPath = path.join(this.outputDir, `p_agent_${type}.md`)
+    await fs.mkdir(this.outputDir, { recursive: true })
+    const outputPath = path.join(this.outputDir, `${type}.md`)
     await fs.writeFile(outputPath, content, 'utf-8')
   }
 
@@ -461,10 +470,10 @@ class AgentGenerator {
     // Check which agents currently exist
     try {
       const files = await fs.readdir(this.outputDir)
-      const agentFiles = files.filter(f => f.startsWith('p_agent_') && f.endsWith('.md'))
+      const agentFiles = files.filter(f => f.endsWith('.md') && !f.startsWith('.'))
 
       for (const file of agentFiles) {
-        const type = file.replace('p_agent_', '').replace('.md', '')
+        const type = file.replace('.md', '')
 
         try {
           await this.generateAgent(type, analysis)
@@ -493,10 +502,10 @@ class AgentGenerator {
 
     try {
       const files = await fs.readdir(this.outputDir)
-      const agentFiles = files.filter(f => f.startsWith('p_agent_') && f.endsWith('.md'))
+      const agentFiles = files.filter(f => f.endsWith('.md') && !f.startsWith('.'))
 
       for (const file of agentFiles) {
-        const type = file.replace('p_agent_', '').replace('.md', '')
+        const type = file.replace('.md', '')
 
         if (!requiredAgents.includes(type)) {
           const filePath = path.join(this.outputDir, file)
@@ -513,4 +522,4 @@ class AgentGenerator {
   }
 }
 
-module.exports = new AgentGenerator()
+module.exports = AgentGenerator