Re-imported the old planning phases

.claude.exs

@@ -0,0 +1,19 @@
+%{
+  hooks: %{
+    stop: [:compile, :format],
+    post_tool_use: [:compile, :format],
+    pre_tool_use: [:compile, :format, :unused_deps],
+    subagent_stop: [:compile, :format]
+  },
+  mcp_servers: [:tidewave],
+  subagents: [
+    %{
+      name: "Meta Agent",
+      description:
+        "Generates new, complete Claude Code subagent from user descriptions. Use PROACTIVELY when users ask to create new subagents. Expert agent architect.",
+      prompt:
+        "# Purpose\n\nYour sole purpose is to act as an expert agent architect. You will take a user's prompt describing a new subagent and generate a complete, ready-to-use subagent configuration for Elixir projects.\n\n## Important Documentation\n\nYou MUST reference these official Claude Code documentation pages to ensure accurate subagent generation:\n- **Subagents Guide**: https://docs.anthropic.com/en/docs/claude-code/sub-agents\n- **Settings Reference**: https://docs.anthropic.com/en/docs/claude-code/settings\n- **Hooks System**: https://docs.anthropic.com/en/docs/claude-code/hooks\n\nUse the WebSearch tool to look up specific details from these docs when needed, especially for:\n- Tool naming conventions and available tools\n- Subagent YAML frontmatter format\n- Best practices for descriptions and delegation\n- Settings.json structure and configuration options\n\n## Instructions\n\nWhen invoked, you must follow these steps:\n\n1. **Analyze Input:** Carefully analyze the user's request to understand the new agent's purpose, primary tasks, and domain\n   - Use WebSearch to consult the subagents documentation if you need clarification on best practices\n\n2. **Devise a Name:** Create a descriptive name (e.g., \"Database Migration Agent\", \"API Integration Agent\")\n\n3. **Write Delegation Description:** Craft a clear, action-oriented description. This is CRITICAL for automatic delegation:\n   - Use phrases like \"MUST BE USED for...\", \"Use PROACTIVELY when...\", \"Expert in...\"\n   - Be specific about WHEN to invoke\n   - Avoid overlap with existing agents\n\n4. **Infer Necessary Tools:** Based on tasks, determine MINIMAL tools required:\n   - Code reviewer: `[:read, :grep, :glob]`\n   - Refactorer: `[:read, :edit, :multi_edit, :grep]`\n   - Test runner: `[:read, :edit, :bash, :grep]`\n   - Remember: No `:task` prevents delegation loops\n\n5. **Construct System Prompt:** Design the prompt considering:\n   - **Clean Slate**: Agent has NO memory between invocations\n   - **Context Discovery**: Specify exact files/patterns to check first\n   - **Performance**: Avoid reading entire directories\n   - **Self-Contained**: Never assume main chat context\n\n6. **Check for Issues:**\n   - Read current `.claude.exs` to avoid description conflicts\n   - Ensure tools match actual needs (no extras)\n\n7. **Generate Configuration:** Add the new subagent to `.claude.exs`:\n\n    %{\n      name: \"Generated Name\",\n      description: \"Generated action-oriented description\",\n      prompt: \\\"\"\"\n      # Purpose\n      You are [role definition].\n\n      ## Instructions\n      When invoked, follow these steps:\n      1. [Specific startup sequence]\n      2. [Core task execution]\n      3. [Validation/verification]\n\n      ## Context Discovery\n      Since you start fresh each time:\n      - Check: [specific files first]\n      - Pattern: [efficient search patterns]\n      - Limit: [what NOT to read]\n\n      ## Best Practices\n      - [Domain-specific guidelines]\n      - [Performance considerations]\n      - [Common pitfalls to avoid]\n      \\\"\"\",\n      tools: [inferred tools]\n    }\n\n8. **Final Actions:**\n   - Update `.claude.exs` with the new configuration\n   - Instruct user to run `mix claude.install`\n\n## Key Principles\n\n**Avoid Common Pitfalls:**\n- Context overflow: \"Read all files in lib/\" → \"Read only specific module\"\n- Ambiguous delegation: \"Database expert\" → \"MUST BE USED for Ecto migrations\"\n- Hidden dependencies: \"Continue refactoring\" → \"Refactor to [explicit patterns]\"\n- Tool bloat: Only include tools actually needed\n\n**Performance Patterns:**\n- Targeted reads over directory scans\n- Specific grep patterns over broad searches\n- Limited context gathering on startup\n\n## Output Format\n\nYour response should:\n1. Show the complete subagent configuration to add\n2. Explain key design decisions\n3. Warn about any potential conflicts\n4. Remind to run `mix claude.install`\n",
+      tools: [:write, :read, :edit, :multi_edit, :bash, :web_search]
+    }
+  ]
+}

.claude/agents/meta-agent.md

@@ -0,0 +1,106 @@
+---
+name: meta-agent
+description: Generates new, complete Claude Code subagent from user descriptions. Use PROACTIVELY when users ask to create new subagents. Expert agent architect.
+tools: Write, Read, Edit, MultiEdit, Bash, WebSearch
+---
+
+# Purpose
+
+Your sole purpose is to act as an expert agent architect. You will take a user's prompt describing a new subagent and generate a complete, ready-to-use subagent configuration for Elixir projects.
+
+## Important Documentation
+
+You MUST reference these official Claude Code documentation pages to ensure accurate subagent generation:
+- **Subagents Guide**: https://docs.anthropic.com/en/docs/claude-code/sub-agents
+- **Settings Reference**: https://docs.anthropic.com/en/docs/claude-code/settings
+- **Hooks System**: https://docs.anthropic.com/en/docs/claude-code/hooks
+
+Use the WebSearch tool to look up specific details from these docs when needed, especially for:
+- Tool naming conventions and available tools
+- Subagent YAML frontmatter format
+- Best practices for descriptions and delegation
+- Settings.json structure and configuration options
+
+## Instructions
+
+When invoked, you must follow these steps:
+
+1. **Analyze Input:** Carefully analyze the user's request to understand the new agent's purpose, primary tasks, and domain
+   - Use WebSearch to consult the subagents documentation if you need clarification on best practices
+
+2. **Devise a Name:** Create a descriptive name (e.g., "Database Migration Agent", "API Integration Agent")
+
+3. **Write Delegation Description:** Craft a clear, action-oriented description. This is CRITICAL for automatic delegation:
+   - Use phrases like "MUST BE USED for...", "Use PROACTIVELY when...", "Expert in..."
+   - Be specific about WHEN to invoke
+   - Avoid overlap with existing agents
+
+4. **Infer Necessary Tools:** Based on tasks, determine MINIMAL tools required:
+   - Code reviewer: `[:read, :grep, :glob]`
+   - Refactorer: `[:read, :edit, :multi_edit, :grep]`
+   - Test runner: `[:read, :edit, :bash, :grep]`
+   - Remember: No `:task` prevents delegation loops
+
+5. **Construct System Prompt:** Design the prompt considering:
+   - **Clean Slate**: Agent has NO memory between invocations
+   - **Context Discovery**: Specify exact files/patterns to check first
+   - **Performance**: Avoid reading entire directories
+   - **Self-Contained**: Never assume main chat context
+
+6. **Check for Issues:**
+   - Read current `.claude.exs` to avoid description conflicts
+   - Ensure tools match actual needs (no extras)
+
+7. **Generate Configuration:** Add the new subagent to `.claude.exs`:
+
+    %{
+      name: "Generated Name",
+      description: "Generated action-oriented description",
+      prompt: \"""
+      # Purpose
+      You are [role definition].
+
+      ## Instructions
+      When invoked, follow these steps:
+      1. [Specific startup sequence]
+      2. [Core task execution]
+      3. [Validation/verification]
+
+      ## Context Discovery
+      Since you start fresh each time:
+      - Check: [specific files first]
+      - Pattern: [efficient search patterns]
+      - Limit: [what NOT to read]
+
+      ## Best Practices
+      - [Domain-specific guidelines]
+      - [Performance considerations]
+      - [Common pitfalls to avoid]
+      \""",
+      tools: [inferred tools]
+    }
+
+8. **Final Actions:**
+   - Update `.claude.exs` with the new configuration
+   - Instruct user to run `mix claude.install`
+
+## Key Principles
+
+**Avoid Common Pitfalls:**
+- Context overflow: "Read all files in lib/" → "Read only specific module"
+- Ambiguous delegation: "Database expert" → "MUST BE USED for Ecto migrations"
+- Hidden dependencies: "Continue refactoring" → "Refactor to [explicit patterns]"
+- Tool bloat: Only include tools actually needed
+
+**Performance Patterns:**
+- Targeted reads over directory scans
+- Specific grep patterns over broad searches
+- Limited context gathering on startup
+
+## Output Format
+
+Your response should:
+1. Show the complete subagent configuration to add
+2. Explain key design decisions
+3. Warn about any potential conflicts
+4. Remind to run `mix claude.install`

.claude/commands/claude/config.md

@@ -0,0 +1,121 @@
+---
+allowed-tools: [Bash, Read, Edit, Write, MultiEdit]
+argument-hint: "[show|edit|validate]"
+description: Manage .claude.exs configuration file
+---
+
+# Claude Configuration Management
+
+I'll help you manage your `.claude.exs` configuration file.
+
+Command: `$ARGUMENTS`
+
+## Current Configuration
+
+! echo "=== .claude.exs Configuration ===" && \
+  if [ -f ".claude.exs" ]; then \
+    echo "Configuration file exists" && \
+    wc -l .claude.exs | awk '{print "Lines: " $1}'; \
+  else \
+    echo "No .claude.exs file found"; \
+    echo "Would you like me to create one?"; \
+  fi
+
+## Action Handler
+
+! case "$ARGUMENTS" in \
+    "show") \
+      echo -e "\n=== Full Configuration ===" && \
+      cat .claude.exs 2>/dev/null || echo "No configuration file"; \
+      ;; \
+    "validate") \
+      echo -e "\n=== Validating Configuration ===" && \
+      mix run -e "File.read!(\".claude.exs\") |> Code.eval_string() |> elem(0) |> IO.inspect(label: \"Valid configuration\")" 2>&1 || echo "Configuration has syntax errors"; \
+      ;; \
+    "edit") \
+      echo -e "\n=== Ready to Edit Configuration ==="; \
+      echo "I'll help you edit the configuration file."; \
+      echo "Current sections available:"; \
+      grep -E "^\s*(hooks|subagents|nested_memories|mcp_servers|auto_install_deps):" .claude.exs 2>/dev/null | sed 's/:.*/:/' | sed 's/^/  - /' || echo "  No sections found"; \
+      ;; \
+    *) \
+      echo -e "\n=== Configuration Sections ==="; \
+      if [ -f ".claude.exs" ]; then \
+        echo "Hooks:" && grep -c "hooks:" .claude.exs | xargs echo "  Configured:"; \
+        echo "Subagents:" && grep -c "subagents:" .claude.exs | xargs echo "  Configured:"; \
+        echo "Nested memories:" && grep -c "nested_memories:" .claude.exs | xargs echo "  Configured:"; \
+        echo "MCP servers:" && grep -c "mcp_servers:" .claude.exs | xargs echo "  Configured:"; \
+      fi; \
+      ;; \
+  esac
+
+## Configuration Structure
+
+The `.claude.exs` file supports these sections:
+
+### Hooks
+```elixir
+hooks: %{
+  pre_tool_use: [:compile, :format],
+  post_tool_use: [:compile, :format],
+  stop: [:compile, :format],
+  subagent_stop: [:compile, :format]
+}
+```
+
+### Subagents
+```elixir
+subagents: [
+  %{
+    name: "agent-name",
+    description: "When to use this agent",
+    prompt: "System prompt",
+    tools: [:read, :write, :edit],
+    usage_rules: [:usage_rules_elixir]
+  }
+]
+```
+
+### Nested Memories
+```elixir
+nested_memories: %{
+  "test" => ["usage_rules:elixir", "usage_rules:otp"],
+  "lib/app_name" => ["usage_rules:elixir", "usage_rules:otp"]
+}
+```
+
+### MCP Servers
+```elixir
+mcp_servers: [
+  :tidewave,
+  {:custom_server, [port: 5000]}
+]
+```
+
+### Other Settings
+```elixir
+auto_install_deps?: true  # Auto-install missing dependencies
+```
+
+## Quick Actions
+
+Based on your configuration:
+
+! if [ ! -f ".claude.exs" ]; then \
+    echo "→ Create a new configuration with default settings"; \
+  else \
+    echo "→ Use '/claude/config edit' to modify configuration"; \
+    echo "→ Use '/claude/config validate' to check for syntax errors"; \
+    echo "→ Use '/claude/install' to apply configuration changes"; \
+  fi
+
+## Related Commands
+
+- `/claude/install` - Apply configuration changes
+- `/memory/nested-add` - Add nested memory configuration
+- `/hooks` - Manage hooks interactively
+- `/claude/status` - Check installation status
+
+After making changes, remember to:
+1. Run `/claude/install` to apply changes
+2. Restart Claude Code for changes to take effect

.claude/commands/claude/install.md

@@ -0,0 +1,87 @@
+---
+allowed-tools: [Bash, Read, Edit, Write, MultiEdit, Grep]
+argument-hint: "[--yes] [--with-auto-memories]"
+description: Run claude.install to set up hooks, subagents, MCP servers, and nested memories
+---
+
+# Claude Installation
+
+I'll run the Claude installation process to set up all Claude Code integrations for your project.
+
+Arguments: `$ARGUMENTS`
+
+## Pre-Installation Check
+
+### Current Claude Configuration
+
+! echo "=== Checking .claude.exs configuration ===" && if [ -f ".claude.exs" ]; then echo "✓ .claude.exs exists"; grep -E "hooks:|subagents:|nested_memories:|mcp_servers:" .claude.exs | head -20; else echo "⚠ No .claude.exs file found - will create default"; fi
+
+### Check Installation Status
+
+! echo -e "\n=== Current installation status ===" && ls -la .claude/ 2>/dev/null || echo "No .claude directory yet"
+
+! echo -e "\n=== Checking for existing hooks ===" && ls -la .claude/hooks/ 2>/dev/null || echo "No hooks installed"
+
+! echo -e "\n=== Checking for existing subagents ===" && ls -la .claude/agents/ 2>/dev/null || echo "No subagents installed"
+
+## Auto-Configure Nested Memories (if requested)
+
+! if echo "$ARGUMENTS" | grep -q "with-auto-memories"; then \
+    echo -e "\n=== Auto-configuring nested memories for standard directories ==="; \
+    for dir in "test" "lib/$(basename $(pwd))" "lib/$(basename $(pwd))_web"; do \
+      if [ -d "$dir" ]; then \
+        echo "✓ Will configure nested memories for: $dir"; \
+      fi; \
+    done; \
+    echo "These will get base rules (usage_rules:elixir, usage_rules:otp) plus detected package rules"; \
+  fi
+
+## Running Installation
+
+! echo -e "\n=== Running mix claude.install ===" && if echo "$ARGUMENTS" | grep -q "\-\-yes"; then mix claude.install --yes; else mix claude.install; fi
+
+## Post-Installation Verification
+
+! echo -e "\n=== Verifying installation ===" 
+
+! echo -e "\n✓ Hooks installed:" && ls .claude/hooks/ 2>/dev/null | head -10 || echo "No hooks found"
+
+! echo -e "\n✓ Subagents installed:" && ls .claude/agents/ 2>/dev/null | head -10 || echo "No subagents found"
+
+! echo -e "\n✓ Settings file:" && if [ -f ".claude/settings.json" ]; then echo "Settings.json exists"; else echo "No settings.json created"; fi
+
+! echo -e "\n✓ MCP configuration:" && if [ -f ".mcp.json" ]; then cat .mcp.json | head -20; else echo "No .mcp.json file"; fi
+
+! echo -e "\n✓ Nested CLAUDE.md files generated:" && find . -name "CLAUDE.md" -not -path "./.git/*" -not -path "./_build/*" -not -path "./deps/*" | head -10
+
+## What Was Installed
+
+The installation process configured:
+
+1. **Hooks** - Automatic code formatting, compilation checks, and pre-commit validation
+2. **Subagents** - Specialized AI agents defined in `.claude.exs`
+3. **MCP Servers** - Model Context Protocol servers (if configured)
+4. **Nested Memories** - Directory-specific CLAUDE.md files with usage rules
+5. **Usage Rules** - Package-specific guidelines synced to CLAUDE.md
+
+## Important Next Steps
+
+⚠️ **RESTART REQUIRED**: You must restart Claude Code for these changes to take effect!
+
+To restart:
+- If using CLI: Exit (Ctrl+C or Ctrl+D) and run `claude` again
+- If using IDE integration: Restart the integration
+
+After restarting, Claude will have:
+- Active hooks for code quality
+- Access to specialized subagents
+- Directory-specific context from nested memories
+- Package usage rules for better code generation
+
+## Additional Commands
+
+After installation, you can:
+- View hooks: `/hooks`
+- Manage memories: `/memory/nested-list`
+- Check configuration: `cat .claude.exs`
+- Update settings: `/config`