oh-my-customcode 0.12.3 → 0.13.0

package/README.md

@@ -26,6 +26,7 @@ Like oh-my-zsh transformed shell customization, oh-my-customcode makes personali
 | **Dead Simple Customization** | Create a folder + markdown file = new agent or skill |
 | **Mix and Match** | Use built-in components, create your own, or combine both |
 | **Non-Destructive** | Your customizations live alongside defaults, never overwritten |
+| **Dynamic Agent Creation** | No matching expert? The system creates one on-the-fly, connecting relevant skills and guides |
 
 ## Quick Start
 

package/dist/cli/index.js

@@ -14074,11 +14074,18 @@ async function generateMCPConfig(targetDir) {
   if (!ontologyExists) {
     return;
   }
+  try {
+    execSync3("uv venv .venv", { cwd: targetDir, stdio: "pipe" });
+    execSync3("uv pip install ontology-rag", { cwd: targetDir, stdio: "pipe" });
+  } catch (error2) {
+    const msg = error2 instanceof Error ? error2.message : String(error2);
+    throw new Error(`Failed to setup Python environment: ${msg}`);
+  }
   const config = {
     mcpServers: {
       "ontology-rag": {
         type: "stdio",
-        command: "python",
+        command: ".venv/bin/python",
         args: ["-m", "ontology_rag.mcp_server"],
         env: {
           ONTOLOGY_DIR: ontologyDir
@@ -14107,9 +14114,9 @@ async function generateMCPConfig(targetDir) {
 `);
   }
 }
-async function checkPythonAvailable() {
+async function checkUvAvailable() {
   try {
-    execSync3("python --version", { stdio: "pipe" });
+    execSync3("uv --version", { stdio: "pipe" });
     return true;
   } catch {
     return false;
@@ -14187,16 +14194,17 @@ async function initCommand(options) {
     const installedPaths = buildInstalledPaths(targetDir, installResult.installedComponents);
     logInstallResultInfo(installResult);
     logSuccessDetails(installedPaths, installResult.skippedComponents);
-    const pythonAvailable = await checkPythonAvailable();
-    if (pythonAvailable) {
+    const uvAvailable = await checkUvAvailable();
+    if (uvAvailable) {
       try {
         await generateMCPConfig(targetDir);
-      } catch {
-        console.warn("Warning: Failed to generate MCP config. You can configure it manually.");
+      } catch (error2) {
+        const msg = error2 instanceof Error ? error2.message : String(error2);
+        console.warn(`Warning: Failed to setup MCP environment: ${msg}`);
       }
     } else {
-      console.warn("Warning: Python not found. Skipping MCP server configuration.");
-      console.warn("Install Python 3.10+ and ontology-rag to enable MCP integration.");
+      console.warn("Warning: uv not found. Skipping MCP server configuration.");
+      console.warn("Install uv (https://docs.astral.sh/uv/) to enable MCP integration.");
     }
     return {
       success: true,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-customcode",
-  "version": "0.12.3",
+  "version": "0.13.0",
   "description": "Batteries-included agent harness for Claude Code",
   "type": "module",
   "bin": {

package/templates/.claude/agents/arch-documenter.md

@@ -4,7 +4,6 @@ description: Use for generating architecture documentation, API specifications (
 model: sonnet
 memory: project
 effort: high
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/arch-speckit-agent.md

@@ -4,7 +4,6 @@ description: Use for spec-driven development, transforming requirements into exe
 model: sonnet
 memory: project
 effort: high
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/be-express-expert.md

@@ -4,7 +4,6 @@ description: Expert Express.js developer for production-ready Node.js APIs follo
 model: sonnet
 memory: project
 effort: high
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/be-nestjs-expert.md

@@ -4,7 +4,6 @@ description: Expert NestJS developer for opinionated, scalable Node.js applicati
 model: sonnet
 memory: project
 effort: high
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/db-supabase-expert.md

@@ -4,6 +4,8 @@ description: Supabase and PostgreSQL expert. Use when working with Supabase proj
 model: sonnet
 memory: user
 effort: high
+skills:
+  - supabase-postgres-best-practices
 tools:
   - Read
   - Write
@@ -11,8 +13,6 @@ tools:
   - Grep
   - Glob
   - Bash
-skills:
-  - supabase-postgres-best-practices
 ---
 
 You are an expert in Supabase and PostgreSQL for performant, secure database-driven applications.

package/templates/.claude/agents/fe-svelte-agent.md

@@ -4,7 +4,6 @@ description: Use for Svelte compiler-based reactivity, reactive statements ($:),
 model: sonnet
 memory: project
 effort: medium
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/fe-vuejs-agent.md

@@ -4,7 +4,6 @@ description: Use for Vue 3 Composition API development, single-file components,
 model: sonnet
 memory: project
 effort: medium
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/lang-java21-expert.md

@@ -4,7 +4,6 @@ description: Expert Java 21 developer for modern Java with Virtual Threads, Patt
 model: sonnet
 memory: project
 effort: high
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/mgr-creator.md

@@ -37,3 +37,18 @@ No registry update needed - agents auto-discovered from `.claude/agents/*.md`.
 
 - R000: All files in English
 - R006: Agent file = role/capabilities only; skills = instructions; guides = reference docs
+
+## Dynamic Creation Mode
+
+When invoked as routing fallback (not explicit `/create-agent`):
+
+1. Receive context: detected domain, keywords, file patterns
+2. Auto-discover: scan `.claude/skills/` for matching skills
+3. Auto-connect: scan `guides/` for relevant reference docs
+4. Create minimal viable agent with:
+   - Detected skills and relevant guides
+   - `sonnet` model (default)
+   - `project` memory scope
+5. Agent is persisted (not ephemeral) for future reuse
+
+Dynamic mode skips user confirmation and creates the agent immediately to fulfill the pending task.

package/templates/.claude/agents/mgr-gitnerd.md

@@ -4,7 +4,6 @@ description: Use when you need to handle Git operations and GitHub workflow mana
 model: sonnet
 memory: project
 effort: medium
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/qa-engineer.md

@@ -4,7 +4,6 @@ description: Use when you need to execute tests based on detailed plans and docu
 model: sonnet
 memory: project
 effort: medium
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/qa-planner.md

@@ -4,7 +4,6 @@ description: Use when you need to create detailed QA plans from requirements and
 model: sonnet
 memory: project
 effort: high
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/qa-writer.md

@@ -4,7 +4,6 @@ description: Use when you need to create comprehensive QA documentation from det
 model: sonnet
 memory: project
 effort: medium
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/sys-naggy.md

@@ -4,7 +4,6 @@ description: Use when you need TODO list management and task tracking with proac
 model: sonnet
 memory: local
 effort: low
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/agents/tool-bun-expert.md

@@ -4,7 +4,6 @@ description: Use for Bun runtime development, bunfig.toml configuration, Bun tes
 model: sonnet
 memory: project
 effort: medium
-skills: []
 tools:
   - Read
   - Write

package/templates/.claude/rules/MUST-continuous-improvement.md

@@ -1,6 +1,6 @@
 # [MUST] Continuous Improvement Rules
 
-> **Priority**: MUST | **ID**: R015 | **Trigger**: User points out rule violation
+> **Priority**: MUST | **ID**: R016 | **Trigger**: User points out rule violation
 
 ## Core Rule
 

package/templates/.claude/rules/MUST-intent-transparency.md

@@ -1,6 +1,6 @@
 # [MUST] Intent Transparency Rules
 
-> **Priority**: MUST | **ID**: R014
+> **Priority**: MUST | **ID**: R015
 
 ## Core Rule
 

package/templates/.claude/rules/MUST-orchestrator-coordination.md

@@ -40,6 +40,7 @@ After restart/compaction: re-read CLAUDE.md, all delegation rules still apply. N
 | Docker/Infra | infra-docker-expert |
 | AWS | infra-aws-expert |
 | Database schema | db-supabase-expert |
+| Unmatched specialized task | mgr-creator → dynamic agent creation |
 
 **Rules:**
 - All file modifications MUST be delegated (orchestrator only uses Read/Glob/Grep)
@@ -62,6 +63,23 @@ Subagent NOT required for:
 
 For specialized work, ALWAYS delegate to appropriate subagent.
 
+## Dynamic Agent Creation (No-Match Fallback)
+
+When routing detects no matching agent for a specialized task:
+
+1. **Evaluate**: Is this a specialized task requiring domain expertise?
+   - YES → proceed to step 2
+   - NO → use general-purpose agent
+2. **Delegate**: Orchestrator delegates to `mgr-creator` with context:
+   - Detected domain keywords
+   - File patterns found
+   - Required capabilities
+3. **Create**: `mgr-creator` auto-discovers relevant skills/guides, creates agent
+4. **Execute**: Orchestrator uses newly created agent for the original task
+
+This is the core oh-my-customcode philosophy:
+> "No expert? CREATE one, connect knowledge, and USE it."
+
 ## Model Selection
 
 ```
@@ -117,7 +135,7 @@ The skill's WORKFLOW is followed, but git EXECUTION is delegated to mgr-gitnerd
 
 ## Agent Teams (when enabled)
 
-When `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`: use Agent Teams for 3+ agent coordinated tasks. See R017 for decision matrix. Task tool remains fallback for simple/independent tasks.
+When `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`: use Agent Teams for 3+ agent coordinated tasks. See R018 for decision matrix. Task tool remains fallback for simple/independent tasks.
 
 ## Announcement Format
 

package/templates/.claude/rules/MUST-sync-verification.md

@@ -1,6 +1,6 @@
 # [MUST] Sync Verification Rules
 
-> **Priority**: MUST - ENFORCED | **ID**: R016
+> **Priority**: MUST - ENFORCED | **ID**: R017
 
 ## Core Rule
 

package/templates/.claude/rules/SHOULD-agent-teams.md

@@ -1,6 +1,6 @@
 # [SHOULD] Agent Teams Rules
 
-> **Priority**: SHOULD | **ID**: R017 | **Condition**: Agent Teams enabled locally
+> **Priority**: SHOULD | **ID**: R018 | **Condition**: Agent Teams enabled locally
 
 ## Detection
 
@@ -8,32 +8,100 @@ Available when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` or TeamCreate/SendMessag
 
 ## Decision Matrix
 
-| Scenario | Preferred |
-|----------|-----------|
-| Simple independent subtasks | Task Tool |
-| Multi-step with shared state | Agent Teams |
-| Research requiring discussion | Agent Teams |
-| Cost-sensitive batch ops | Task Tool |
-| Complex debugging across modules | Agent Teams |
-| Code review + fix cycle | Agent Teams |
-| Single file operations | Task Tool |
+| Scenario | Preferred | Reason |
+|----------|-----------|--------|
+| Simple independent subtasks | Task Tool | Lower cost, no coordination overhead |
+| Multi-step with shared state | **Agent Teams** | Shared task list, peer messaging |
+| Research requiring discussion | **Agent Teams** | Iterative discovery, synthesis |
+| Cost-sensitive batch ops | Task Tool | Minimal token overhead |
+| Complex debugging across modules | **Agent Teams** | Cross-module state sharing |
+| Code review + fix cycle | **Agent Teams** | Review → fix → re-review loop |
+| Single file operations | Task Tool | Overkill for simple tasks |
+| Dynamic agent creation + usage | **Agent Teams** | Create → test → iterate cycle |
 
-**Self-check before Task tool:** Is Agent Teams available? Does task need 3+ agents, shared state, or inter-agent communication? If yes, use Agent Teams.
+**ACTIVELY prefer Agent Teams for qualifying tasks.** The coordination richness justifies the token cost for complex work.
+
+## Self-Check (Mandatory Before Task Tool)
+
+BEFORE using Task tool for 2+ agent tasks, CHECK:
+
+```
+╔════════════════════════════════════════════════════════╗
+║  AGENT TEAMS CHECK:                                    ║
+║                                                        ║
+║  1. Is Agent Teams available?                          ║
+║  2. Does task need 3+ agents?                          ║
+║  3. Do agents need shared state?                       ║
+║  4. Is inter-agent communication needed?               ║
+║  5. Is there a review/fix/iterate cycle?               ║
+║                                                        ║
+║  If YES to #1 AND any of #2-#5 → USE Agent Teams      ║
+╚════════════════════════════════════════════════════════╝
+```
 
 ## Team Patterns
 
+### Standard Patterns
+
 - **Research**: researcher-1 + researcher-2 + synthesizer
 - **Development**: implementer + reviewer + tester
 - **Debug**: investigator-1 + investigator-2 + fixer
 
+### Hybrid Patterns
+
+- **Review+Fix**: reviewer + implementer (reviewer finds issues → implementer fixes → reviewer re-checks)
+- **Create+Validate**: mgr-creator + qa-engineer (create agent → validate → iterate)
+- **Multi-Expert**: expert-1 + expert-2 + coordinator (cross-domain tasks requiring multiple specialties)
+
+### Dynamic Patterns
+
+- **Dynamic Creation**: mgr-creator + domain-expert (create new agent → immediately use for pending task)
+- **Codex Hybrid**: codex-exec-agent + claude-reviewer (Codex generates → Claude reviews/refines)
+
+## Codex-Exec Integration
+
+When both Agent Teams and codex-exec are available:
+
+```
+Hybrid Workflow:
+  1. Claude agent analyzes requirements
+  2. codex-exec generates implementation (Codex strength: code generation)
+  3. Claude agent reviews and refines (Claude strength: reasoning, quality)
+  4. Iterate via team messaging until quality meets standards
+```
+
+| Step | Agent | Model |
+|------|-------|-------|
+| Analysis | Claude team member | sonnet/opus |
+| Generation | codex-exec | o3/o4-mini |
+| Review | Claude team member | sonnet |
+| Refinement | Appropriate expert | sonnet |
+
+## Dynamic Agent Creation in Teams
+
+When Agent Teams creates a new agent via mgr-creator:
+
+1. Team lead identifies missing expertise
+2. Spawns mgr-creator as team member
+3. mgr-creator creates agent with auto-discovered skills
+4. New agent joins team immediately
+5. Team continues with expanded capabilities
+
 ## Lifecycle
 
-TeamCreate → TaskCreate → Task(spawn) → SendMessage(coordinate) → TaskUpdate → shutdown → cleanup
+```
+TeamCreate → TaskCreate → Task(spawn members) → SendMessage(coordinate)
+  → TaskUpdate(progress) → ... → shutdown members → TeamDelete
+```
 
 ## Fallback
 
 When Agent Teams unavailable: use Task tool with R009/R010 rules. Both approaches produce results; Agent Teams adds coordination richness.
 
-## Cost
+## Cost Awareness
 
-Agent Teams uses more tokens (full context per member + message passing). Use Task tool if task < 3 min or no inter-agent communication needed.
+Agent Teams uses more tokens (full context per member + message passing). Use Task tool when:
+- Task takes < 3 minutes
+- No inter-agent communication needed
+- Simple independent subtasks only
+- Cost is the primary concern

package/templates/.claude/rules/index.yaml

@@ -62,7 +62,7 @@ rules:
     scope: agents
 
   # Intent Transparency - MUST
-  - id: R014
+  - id: R015
     name: intent-transparency
     title: Intent Transparency Rules
     path: ./MUST-intent-transparency.md
@@ -70,7 +70,7 @@ rules:
     scope: orchestrator
 
   # Continuous Improvement - MUST
-  - id: R015
+  - id: R016
     name: continuous-improvement
     title: Continuous Improvement Rules
     path: ./MUST-continuous-improvement.md
@@ -78,7 +78,7 @@ rules:
     scope: all
 
   # Sync Verification - MUST
-  - id: R016
+  - id: R017
     name: sync-verification
     title: Sync Verification Rules
     path: ./MUST-sync-verification.md
@@ -125,7 +125,7 @@ rules:
     scope: all
 
   # Agent Teams - SHOULD
-  - id: R017
+  - id: R018
     name: agent-teams
     title: Agent Teams Rules
     path: ./SHOULD-agent-teams.md

package/templates/.claude/skills/codex-exec/SKILL.md

@@ -2,7 +2,6 @@
 name: codex-exec
 description: Execute OpenAI Codex CLI prompts and return results
 argument-hint: "<prompt> [--json] [--output <path>] [--model <name>] [--timeout <ms>]"
-disable-model-invocation: true
 ---
 
 # Codex Exec Skill
@@ -121,3 +120,30 @@ Works with the orchestrator pattern:
 - Main conversation delegates Codex execution via this skill
 - Results are returned to the main conversation for further processing
 - Can be chained with other skills (e.g., dev-review after Codex generates code)
+
+## Availability Check
+
+codex-exec requires the Codex CLI binary to be installed and authenticated. The skill is only usable when:
+
+1. `codex` binary is found in PATH (`which codex` succeeds)
+2. Authentication is valid (OPENAI_API_KEY set or `codex` logged in)
+
+If either check fails, this skill cannot be used. Fall back to Claude agents for the task.
+
+> **Note**: `disable-model-invocation: true` is set intentionally. This skill must be explicitly invoked via `/codex-exec` or delegated by the orchestrator. It is NOT auto-invoked by the model.
+
+## Agent Teams Integration
+
+When used within Agent Teams (requires explicit invocation):
+
+1. **As delegated task**: orchestrator explicitly delegates codex-exec for code generation
+2. **Hybrid workflow**: Claude team member analyzes → orchestrator invokes codex-exec → Claude reviews
+3. **Iteration**: Team messaging enables review-fix cycles between Claude and Codex outputs
+
+```
+Orchestrator delegates generation task
+  → /codex-exec invoked explicitly
+  → Output returned to orchestrator
+  → Reviewer validates quality
+  → Iterate if needed
+```

package/templates/.claude/skills/create-agent/SKILL.md

@@ -23,6 +23,7 @@ Create a new agent with complete directory structure, files, and registration.
 --source, -s     External source URL (for external agents)
 --desc, -d       Description
 --skills         Comma-separated skills to include
+--dynamic        Auto-discover skills and guides from context (used by routing fallback)
 ```
 
 ## Workflow

package/templates/.claude/skills/de-lead-routing/SKILL.md

@@ -219,6 +219,29 @@ Pipeline design completed.
 - **secretary-routing**: DE agents accessible through secretary for management tasks
 - **qa-lead-routing**: Coordinates with QA for data quality testing
 
+## No Match Fallback
+
+When a data engineering tool is detected but no matching agent exists:
+
+```
+User Input → No matching DE agent
+  ↓
+Detect: DE tool keyword or config file pattern
+  ↓
+Delegate to mgr-creator with context:
+  domain: detected DE tool
+  type: de-engineer
+  keywords: extracted tool names
+  file_patterns: detected config patterns
+  skills: auto-discover from .claude/skills/
+  guides: auto-discover from guides/
+```
+
+**Examples of dynamic creation triggers:**
+- New data tools (e.g., "Dagster DAG 만들어줘", "Flink 스트리밍 설정해줘")
+- Unfamiliar data formats or connectors
+- Data tool detected in project but no specialist agent
+
 ## Agent Teams Awareness
 
 Before routing via Task tool, check if Agent Teams is available (`CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` or TeamCreate/SendMessage tools present).

package/templates/.claude/skills/dev-lead-routing/SKILL.md

@@ -77,6 +77,29 @@ user-invocable: false
 
 Multi-language: detect all languages, route to parallel experts (max 4). Single-language: route to matching expert. Cross-layer (frontend + backend): multiple experts in parallel.
 
+## No Match Fallback
+
+When file extension or keyword doesn't match any existing agent:
+
+```
+User Input → No matching development agent
+  ↓
+Detect: File extension (.rb, .swift, .dart, etc.) or language keyword
+  ↓
+Delegate to mgr-creator with context:
+  domain: detected language/framework
+  type: sw-engineer
+  keywords: extracted from user input
+  file_patterns: detected extensions
+  skills: auto-discover from .claude/skills/
+  guides: auto-discover from guides/
+```
+
+**Examples of dynamic creation triggers:**
+- Unrecognized file extension (e.g., `.rb` → Ruby expert, `.swift` → Swift expert)
+- New framework keyword (e.g., "Flutter 앱 리뷰해줘", "Rails API 만들어줘")
+- Language detected but no specialist exists
+
 ## Agent Teams Awareness
 
 Before routing via Task tool, check if Agent Teams is available (`CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` or TeamCreate/SendMessage tools present).

package/templates/.claude/skills/intent-detection/SKILL.md

@@ -176,8 +176,9 @@ Each agent defines:
 
 ## Error Handling
 
-### No Match
+### No Match (< 30% confidence)
 
+**Generic task** (no identifiable domain):
 ```
 [Intent Unclear]
 ├── Input: "도와줘"
@@ -190,6 +191,20 @@ How can I help? Please be more specific:
 - What file or project?
 ```
 
+**Specialized task with identifiable domain** (keywords/files detected but no matching agent):
+```
+[No Matching Agent]
+├── Input: "Terraform 모듈 리뷰해줘"
+├── Domain: terraform (detected from keywords)
+├── Matching Agent: none
+└── Action: Trigger dynamic agent creation
+
+→ Delegating to mgr-creator with context:
+  domain: terraform
+  keywords: ["terraform", "모듈"]
+  file_patterns: ["*.tf"]
+```
+
 ### Ambiguous Match
 
 ```

package/templates/.claude/skills/qa-lead-routing/SKILL.md

@@ -265,6 +265,28 @@ metrics:
   test_case_count: number
 ```
 
+## No Match Fallback
+
+When a QA task involves unfamiliar testing patterns or tools:
+
+```
+User Input → QA task with unrecognized tool/pattern
+  ↓
+Detect: Testing framework or QA methodology keyword
+  ↓
+Delegate to mgr-creator with context:
+  domain: detected QA tool/methodology
+  type: qa-engineer
+  keywords: extracted testing terms
+  skills: auto-discover from .claude/skills/
+  guides: auto-discover from guides/
+```
+
+**Examples of dynamic creation triggers:**
+- New testing frameworks (e.g., "Cypress E2E 테스트 작성해줘", "k6 부하 테스트 설계해줘")
+- Specialized QA methodologies (e.g., "뮤테이션 테스트 전략 만들어줘")
+- Performance/security testing tools not covered by existing agents
+
 ## Agent Teams Awareness
 
 Before routing via Task tool, check if Agent Teams is available (`CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` or TeamCreate/SendMessage tools present).

package/templates/.claude/skills/secretary-routing/SKILL.md

@@ -92,6 +92,26 @@ Before routing via Task tool, check if Agent Teams is available (`CLAUDE_CODE_EX
 | Multi-round verification (sauron) | Task Tool |
 | Agent audit + fix cycle | Agent Teams |
 
+## No Match Fallback
+
+When no manager agent matches the request but the task is clearly management-related:
+
+```
+User Input → No matching manager agent
+  ↓
+Evaluate: Is this a specialized management/tooling task?
+  YES → Delegate to mgr-creator with context:
+        domain: detected tool/technology
+        type: manager
+        skills: auto-discover from .claude/skills/
+  NO  → Ask user for clarification
+```
+
+**Examples of dynamic creation triggers:**
+- New CI/CD tool management (e.g., "ArgoCD 배포 관리해줘")
+- New monitoring tool setup (e.g., "Grafana 대시보드 관리")
+- Unfamiliar package manager operations
+
 ## Usage
 
 This skill is NOT user-invocable. It is automatically triggered when the main conversation detects agent management intent.

package/templates/CLAUDE.md.en

@@ -117,8 +117,8 @@ Flow:
 | R008 | Tool Identification | **ENFORCED** - Display agent when using ANY tool |
 | R009 | Parallel Execution | **ENFORCED** - Parallel execution, large task decomposition |
 | R010 | Orchestrator Coordination | **ENFORCED** - Orchestrator coordination, session continuity, direct action prohibition |
-| R015 | Continuous Improvement | **ENFORCED** - Update rules when violations occur |
-| R016 | Sync Verification | **ENFORCED** - Verify sync before structural changes |
+| R016 | Continuous Improvement | **ENFORCED** - Update rules when violations occur |
+| R017 | Sync Verification | **ENFORCED** - Verify sync before structural changes |
 
 ### SHOULD (Strongly recommended)
 | ID | Rule | Description |
@@ -128,8 +128,8 @@ Flow:
 | R011 | Memory Integration | Session persistence with claude-mem |
 | R012 | HUD Statusline | Real-time status display |
 | R013 | Ecomode | Token efficiency for batch ops |
-| R014 | Intent Transparency | **MUST** - Transparent agent routing |
-| R017 | Agent Teams | Active use of Agent Teams when enabled |
+| R015 | Intent Transparency | **MUST** - Transparent agent routing |
+| R018 | Agent Teams | Active use of Agent Teams when enabled |
 
 ### MAY (Optional)
 | ID | Rule | Description |
@@ -159,7 +159,7 @@ Flow:
 | `/optimize-analyze` | Analyze bundle and performance |
 | `/optimize-bundle` | Optimize bundle size |
 | `/optimize-report` | Generate optimization report |
-| `/sauron-watch` | Full R016 verification |
+| `/sauron-watch` | Full R017 verification |
 | `/lists` | Show all available commands |
 | `/status` | Show system status |
 | `/help` | Show help information |
@@ -171,8 +171,8 @@ project/
 +-- CLAUDE.md                    # Entry point
 +-- .claude/
 |   +-- agents/                  # Subagent definitions (42 files)
-|   +-- skills/                  # Skills (51 directories)
-|   +-- rules/                   # Global rules (R000-R017)
+|   +-- skills/                  # Skills (52 directories)
+|   +-- rules/                   # Global rules (R000-R018)
 |   +-- hooks/                   # Hook scripts (memory, HUD)
 |   +-- contexts/                # Context files (ecomode)
 +-- guides/                      # Reference docs (22 topics)
@@ -188,6 +188,17 @@ Orchestration is handled by routing skills in the main conversation:
 
 The main conversation acts as the sole orchestrator. Subagents cannot spawn other subagents.
 
+### Dynamic Agent Creation
+
+When no existing agent matches a specialized task, the system automatically creates one:
+
+1. Routing skill detects no matching expert
+2. Orchestrator delegates to mgr-creator with detected context
+3. mgr-creator auto-discovers relevant skills and guides
+4. New agent is created and used immediately
+
+This is the core oh-my-customcode philosophy: **"No expert? CREATE one, connect knowledge, and USE it."**
+
 ## Agents Summary
 
 | Type | Count | Agents |
@@ -216,8 +227,9 @@ When Claude Code's Agent Teams feature is enabled (`CLAUDE_CODE_EXPERIMENTAL_AGE
 | Best for | Focused tasks | Research, review, debugging |
 | Token cost | Lower | Higher |
 
-**When enabled, actively use Agent Teams for complex collaborative tasks.**
-See R017 (SHOULD-agent-teams.md) for the decision matrix.
+**When enabled, ACTIVELY prefer Agent Teams for qualifying collaborative tasks.**
+See R018 (SHOULD-agent-teams.md) for the decision matrix.
+Hybrid patterns (Claude + Codex, Dynamic Creation + Teams) are supported.
 Task tool + routing skills remain the fallback for simple/cost-sensitive tasks.
 
 ## Quick Reference

package/templates/CLAUDE.md.ko

@@ -117,8 +117,8 @@ oh-my-customcode로 구동됩니다.
 | R008 | 도구 식별 | **강제** - 모든 도구 사용 시 에이전트 표시 |
 | R009 | 병렬 실행 | **강제** - 병렬 실행, 대규모 작업 분해 |
 | R010 | 오케스트레이터 조율 | **강제** - 오케스트레이터 조율, 세션 연속성, 직접 실행 금지 |
-| R015 | 지속적 개선 | **강제** - 위반 발생 시 규칙 업데이트 |
-| R016 | 동기화 검증 | **강제** - 구조 변경 전 검증 |
+| R016 | 지속적 개선 | **강제** - 위반 발생 시 규칙 업데이트 |
+| R017 | 동기화 검증 | **강제** - 구조 변경 전 검증 |
 
 ### SHOULD (강력 권장)
 | ID | 규칙 | 설명 |
@@ -128,8 +128,8 @@ oh-my-customcode로 구동됩니다.
 | R011 | 메모리 통합 | claude-mem을 통한 세션 지속성 |
 | R012 | HUD 상태줄 | 실시간 상태 표시 |
 | R013 | Ecomode | 배치 작업 토큰 효율성 |
-| R014 | 의도 투명성 | **필수** - 투명한 에이전트 라우팅 |
-| R017 | Agent Teams | 활성화 시 Agent Teams 적극 사용 |
+| R015 | 의도 투명성 | **필수** - 투명한 에이전트 라우팅 |
+| R018 | Agent Teams | 활성화 시 Agent Teams 적극 사용 |
 
 ### MAY (선택)
 | ID | 규칙 | 설명 |
@@ -159,7 +159,7 @@ oh-my-customcode로 구동됩니다.
 | `/optimize-analyze` | 번들 및 성능 분석 |
 | `/optimize-bundle` | 번들 크기 최적화 |
 | `/optimize-report` | 최적화 리포트 생성 |
-| `/sauron-watch` | 전체 R016 검증 |
+| `/sauron-watch` | 전체 R017 검증 |
 | `/lists` | 모든 사용 가능한 커맨드 표시 |
 | `/status` | 시스템 상태 표시 |
 | `/help` | 도움말 표시 |
@@ -171,8 +171,8 @@ project/
 +-- CLAUDE.md                    # 진입점
 +-- .claude/
 |   +-- agents/                  # 서브에이전트 정의 (42 파일)
-|   +-- skills/                  # 스킬 (51 디렉토리)
-|   +-- rules/                   # 전역 규칙 (R000-R017)
+|   +-- skills/                  # 스킬 (52 디렉토리)
+|   +-- rules/                   # 전역 규칙 (R000-R018)
 |   +-- hooks/                   # 훅 스크립트 (메모리, HUD)
 |   +-- contexts/                # 컨텍스트 파일 (ecomode)
 +-- guides/                      # 레퍼런스 문서 (22 토픽)
@@ -188,6 +188,17 @@ project/
 
 메인 대화가 유일한 오케스트레이터 역할을 합니다. 서브에이전트는 다른 서브에이전트를 생성할 수 없습니다.
 
+### 동적 에이전트 생성
+
+기존 에이전트 중 작업에 맞는 전문가가 없으면 자동으로 생성합니다:
+
+1. 라우팅 스킬이 매칭 전문가 없음을 감지
+2. 오케스트레이터가 mgr-creator에 컨텍스트와 함께 위임
+3. mgr-creator가 관련 skills/guides를 자동 탐색
+4. 새 에이전트 생성 후 즉시 사용
+
+이것이 oh-my-customcode의 핵심 철학입니다: **"전문가가 없으면? 만들고, 지식을 연결하고, 사용한다."**
+
 ## 에이전트 요약
 
 | 타입 | 수량 | 에이전트 |
@@ -216,8 +227,9 @@ Claude Code의 Agent Teams 기능이 활성화되어 있으면 (`CLAUDE_CODE_EXP
 | 적합한 작업 | 집중된 작업 | 리서치, 리뷰, 디버깅 |
 | 토큰 비용 | 낮음 | 높음 |
 
-**활성화 시, 복잡한 협업 작업에 Agent Teams를 적극 사용합니다.**
-결정 매트릭스는 R017 (SHOULD-agent-teams.md)을 참조하세요.
+**활성화 시, 적격한 협업 작업에 Agent Teams를 적극적으로 우선 사용합니다.**
+결정 매트릭스는 R018 (SHOULD-agent-teams.md)을 참조하세요.
+하이브리드 패턴 (Claude + Codex, 동적 생성 + Teams)이 지원됩니다.
 단순/비용 민감 작업에는 Task tool + 라우팅 스킬이 폴백으로 유지됩니다.
 
 ## 빠른 참조

package/templates/manifest.json

@@ -1,6 +1,6 @@
 {
   "version": "0.2.0",
-  "lastUpdated": "2026-02-08T07:00:00.000Z",
+  "lastUpdated": "2026-02-17T07:30:00.000Z",
   "components": [
     {
       "name": "rules",
@@ -36,7 +36,7 @@
       "name": "contexts",
       "path": ".claude/contexts",
       "description": "Context configuration files",
-      "files": 1
+      "files": 4
     }
   ],
   "source": "https://github.com/baekenough/oh-my-customcode"