@codemieai/code 0.0.3 → 0.0.4

package/.claude/agents/README.md

@@ -1,298 +0,0 @@
-# Specialized Agents
-
-This directory contains specialized subagent prompts for complex, multi-step workflows in the CodeMie Code project.
-
-## What Are Subagents?
-
-Subagents are specialized AI assistants with focused system prompts for specific tasks. Each agent has a dedicated markdown file containing:
-- Role and responsibilities
-- Available tools and capabilities
-- Step-by-step workflow
-- Error handling strategies
-- Best practices and examples
-
-## Available Agents
-
-### 🚀 Release Manager (`release-manager.md`)
-
-**Purpose:** Automate the complete release process from change analysis to npm publication.
-
-**Capabilities:**
-- Analyze git history and categorize changes
-- Generate structured release notes (Keep a Changelog format)
-- Update package version (semver)
-- Create git tags and commits
-- Create GitHub releases
-- Trigger npm publish workflows
-
-**Trigger Phrases:**
-- "Release version 0.0.2"
-- "Create a new release"
-- "Release a patch/minor/major version"
-- "Use release manager to..."
-- "Prepare a release"
-
-**Example:**
-```
-User: "Release version 0.1.0"
-→ Agent analyzes changes, generates notes, creates release
-```
-
-## How to Use Subagents
-
-### Method 1: Natural Language (Recommended)
-
-Simply use the trigger phrases in your conversation with Claude:
-
-```
-You: "Release version 0.2.0"
-Claude: [Reads release-manager.md and executes the workflow]
-```
-
-### Method 2: Explicit Invocation
-
-Reference the agent directly:
-
-```
-You: "Use the release manager to create version 0.2.0"
-Claude: [Loads and follows release-manager.md instructions]
-```
-
-### Method 3: Manual Execution
-
-Read the agent prompt yourself and ask Claude to follow it:
-
-```
-You: "Follow the release-manager agent workflow for version 0.2.0"
-Claude: [Executes according to the documented workflow]
-```
-
-## How Subagents Work
-
-When you trigger a subagent:
-
-1. **Claude identifies** the intent matches a subagent
-2. **Reads the prompt** from `.claude/agents/{agent-name}.md`
-3. **Follows the workflow** step-by-step as documented
-4. **Tracks progress** using TodoWrite tool
-5. **Reports results** when complete or if intervention needed
-
-## Creating Your Own Subagent
-
-### Step 1: Define the Agent
-
-Create a new markdown file: `.claude/agents/{role}-{function}.md`
-
-Example: `code-reviewer.md`, `test-generator.md`, `security-auditor.md`
-
-### Step 2: Write the System Prompt
-
-Use this template:
-
-```markdown
-# {Agent Name}
-
-You are a specialized {role} agent for the CodeMie Code project. Your job is to {primary goal}.
-
-## Your Capabilities
-
-You have access to these tools:
-- Tool 1: Purpose
-- Tool 2: Purpose
-
-## Workflow
-
-When the user requests {action}, follow these steps:
-
-### Step 1: {Step Name}
-[Detailed instructions]
-
-### Step 2: {Step Name}
-[Detailed instructions]
-
-## Error Handling
-
-[How to handle failures]
-
-## Best Practices
-
-[Guidelines for the agent]
-
-## Examples
-
-[Example interactions]
-```
-
-### Step 3: Document Trigger Phrases
-
-Add your agent to this README with:
-- Purpose description
-- Capabilities list
-- Trigger phrases
-- Example usage
-
-### Step 4: Update CLAUDE.md
-
-Add your agent to the "Specialized Agents" section in CLAUDE.md so Claude knows when to invoke it.
-
-## Agent Design Best Practices
-
-### 1. Clear Workflow
-- Break complex tasks into numbered steps
-- Each step should be atomic and verifiable
-- Use checkboxes for validation criteria
-
-### 2. Error Handling
-- Anticipate failure points
-- Provide clear error messages
-- Suggest recovery actions
-- Allow graceful fallbacks
-
-### 3. User Confirmation
-- Ask before destructive operations
-- Show previews of changes
-- Allow dry-run modes
-- Support rollback if possible
-
-### 4. Progress Tracking
-- Use TodoWrite for multi-step workflows
-- Report what you're doing in real-time
-- Show completion status clearly
-
-### 5. Transparency
-- Log commands being executed
-- Explain why each step is necessary
-- Show intermediate results
-- Provide links and references
-
-### 6. Flexibility
-- Support multiple input formats
-- Handle edge cases gracefully
-- Allow customization via config
-- Provide manual fallback options
-
-## Naming Conventions
-
-**File naming:** `{role}-{function}.md`
-
-Examples:
-- ✅ `release-manager.md`
-- ✅ `code-reviewer.md`
-- ✅ `test-generator.md`
-- ✅ `security-auditor.md`
-- ✅ `dependency-updater.md`
-- ❌ `release.md` (too generic)
-- ❌ `manager.md` (unclear purpose)
-- ❌ `agent1.md` (meaningless)
-
-**Role-function pattern:**
-- **Role**: What the agent is (manager, reviewer, generator, auditor)
-- **Function**: What domain it works in (release, code, test, security)
-
-## Agent Communication
-
-### Input Formats
-
-Agents should accept multiple input styles:
-
-```
-# Explicit
-"Release version 0.2.0"
-
-# Semantic
-"Release a minor version"
-
-# Contextual
-"Create a new release"
-
-# With options
-"Release 0.2.0 as a beta"
-```
-
-### Output Format
-
-Agents should provide structured output:
-
-```
-🎯 Task: [What the agent is doing]
-
-📊 Status: [Current progress]
-
-✅ Completed:
-- Step 1
-- Step 2
-
-⏳ In Progress:
-- Step 3
-
-❌ Failed:
-- Step 4 (reason: ...)
-
-🔜 Pending:
-- Step 5
-- Step 6
-```
-
-## Future Agent Ideas
-
-Here are some agents we might create in the future:
-
-- **`code-reviewer.md`**: Automated code review with best practices
-- **`test-generator.md`**: Generate tests for new code
-- **`security-auditor.md`**: Scan for security vulnerabilities
-- **`dependency-updater.md`**: Update and test dependencies
-- **`migration-helper.md`**: Assist with breaking change migrations
-- **`documentation-writer.md`**: Generate comprehensive docs
-- **`performance-analyzer.md`**: Profile and optimize code
-- **`changelog-maintainer.md`**: Keep CHANGELOG.md up to date
-
-## Integration with CI/CD
-
-Subagent workflows can be extracted into GitHub Actions:
-
-```yaml
-# .github/workflows/release.yml
-name: Release
-on:
-  workflow_dispatch:
-    inputs:
-      version:
-        required: true
-
-jobs:
-  release:
-    runs-on: ubuntu-latest
-    steps:
-      # Follow steps from rm.md
-      - name: Analyze Changes
-        run: ...
-      - name: Generate Notes
-        run: ...
-      - name: Create Release
-        run: ...
-```
-
-This allows both manual (via Claude) and automated (via CI) releases following the same workflow.
-
-## Contributing
-
-When adding a new agent:
-
-1. Create the agent file in `.claude/agents/`
-2. Follow the template and best practices
-3. Add documentation to this README
-4. Update CLAUDE.md with trigger phrases
-5. Test the agent with real scenarios
-6. Commit with message: `feat: add {agent-name} subagent`
-
-## Support
-
-For questions about subagents:
-- Review existing agents as examples
-- Check CLAUDE.md for integration details
-- Consult the main README.md for project context
-- Test iteratively with dry runs
-
----
-
-**Remember:** Subagents are here to make complex workflows simple, safe, and repeatable!