fix(post-update): clear Claude plugin cache on update

The Claude plugin cache at ~/.claude/plugins/cache/leo-claude-mktplace/
holds versioned copies of .mcp.json files. When we update .mcp.json
to use run.sh instead of direct python paths, the cache retains old
versions, causing MCP servers to fail.

Now post-update.sh clears this cache, forcing Claude to read fresh
configs from the installed marketplace on next session start.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

.claude-plugin/marketplace.json

@@ -6,7 +6,7 @@
   },
   "metadata": {
     "description": "Project management plugins with Gitea and NetBox integrations",
-    "version": "5.0.0"
+    "version": "5.1.0"
   },
   "plugins": [
     {
@@ -75,8 +75,8 @@
     },
     {
       "name": "cmdb-assistant",
-      "version": "1.0.0",
-      "description": "NetBox CMDB integration for infrastructure management",
+      "version": "1.1.0",
+      "description": "NetBox CMDB integration with data quality validation and machine registration",
       "source": "./plugins/cmdb-assistant",
       "author": {
         "name": "Leo Miranda",
@@ -86,7 +86,7 @@
       "repository": "https://gitea.hotserv.cloud/personal-projects/leo-claude-mktplace.git",
       "mcpServers": ["./.mcp.json"],
       "category": "infrastructure",
-      "tags": ["cmdb", "netbox", "dcim", "ipam"],
+      "tags": ["cmdb", "netbox", "dcim", "ipam", "data-quality", "validation"],
       "license": "MIT"
     },
     {

.gitignore

@@ -31,6 +31,8 @@ venv/
 ENV/
 env/
 .venv/
+.venv
+**/.venv
 
 # PyCharm
 .idea/

CHANGELOG.md

@@ -4,6 +4,51 @@ All notable changes to the Leo Claude Marketplace will be documented in this fil
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 
+## [Unreleased]
+
+### Added
+
+#### Gitea MCP Server - create_pull_request Tool
+- **`create_pull_request`**: Create new pull requests via MCP
+  - Parameters: title, body, head (source branch), base (target branch), labels
+  - Branch-aware security: only allowed on development/feature branches
+  - Completes the PR lifecycle (was previously missing - only had list/get/review/comment)
+
+#### cmdb-assistant v1.1.0 - Data Quality Validation
+- **SessionStart Hook**: Tests NetBox API connectivity at session start
+  - Warns if VMs exist without site assignment
+  - Warns if devices exist without platform
+  - Non-blocking: displays warning, doesn't prevent work
+- **PreToolUse Hook**: Validates input parameters before VM/device operations
+  - Warns about missing site, tenant, platform
+  - Non-blocking: suggests best practices without blocking
+- **`/cmdb-audit` Command**: Comprehensive data quality analysis
+  - Scopes: all, vms, devices, naming, roles
+  - Identifies Critical/High/Medium/Low issues
+  - Provides prioritized remediation recommendations
+- **`/cmdb-register` Command**: Register current machine into NetBox
+  - Discovers system info: hostname, platform, hardware, network interfaces
+  - Discovers running apps: Docker containers, systemd services
+  - Creates device with interfaces, IPs, and sets primary IP
+  - Creates cluster and VMs for Docker containers
+- **`/cmdb-sync` Command**: Sync machine state with NetBox
+  - Compares current state with NetBox record
+  - Shows diff of changes (interfaces, IPs, containers)
+  - Updates with user confirmation
+  - Supports --full and --dry-run flags
+- **NetBox Best Practices Skill**: Reference documentation
+  - Dependency order for object creation
+  - Naming conventions (`{role}-{site}-{number}`, `{env}-{app}-{number}`)
+  - Role consolidation guidance
+  - Site/tenant/platform assignment requirements
+- **Agent Enhancement**: Updated cmdb-assistant agent with validation requirements
+  - Proactive suggestions for missing fields
+  - Naming convention checks
+  - Dependency order enforcement
+  - Duplicate prevention
+
+---
+
 ## [5.0.0] - 2026-01-26
 
 ### Added

README.md

@@ -1,4 +1,4 @@
-# Leo Claude Marketplace - v5.0.0
+# Leo Claude Marketplace - v5.1.0
 
 A collection of Claude Code plugins for project management, infrastructure automation, and development workflows.
 
@@ -53,6 +53,19 @@ Analyze, optimize, and create CLAUDE.md configuration files for Claude Code proj
 
 **Commands:** `/config-analyze`, `/config-optimize`, `/config-init`
 
+#### [contract-validator](./plugins/contract-validator/README.md) *NEW in v5.0.0*
+**Cross-Plugin Compatibility Validation**
+
+Validate plugin marketplaces for command conflicts, tool overlaps, and broken agent references.
+
+- Interface parsing from plugin README.md files
+- Agent extraction from CLAUDE.md definitions
+- Pairwise compatibility checks between all plugins
+- Data flow validation for agent sequences
+- Markdown or JSON reports with actionable suggestions
+
+**Commands:** `/validate-contracts`, `/check-agent`, `/list-interfaces`, `/initial-setup`
+
 ### Productivity
 
 #### [clarity-assist](./plugins/clarity-assist/README.md) *NEW in v3.0.0*
@@ -98,7 +111,7 @@ Full CRUD operations for network infrastructure management directly from Claude
 
 ### Data Engineering
 
-#### [data-platform](./plugins/data-platform/README.md) *NEW*
+#### [data-platform](./plugins/data-platform/README.md) *NEW in v4.0.0*
 **pandas, PostgreSQL/PostGIS, and dbt Integration**
 
 Comprehensive data engineering toolkit with persistent DataFrame storage.
@@ -113,7 +126,7 @@ Comprehensive data engineering toolkit with persistent DataFrame storage.
 
 ### Visualization
 
-#### [viz-platform](./plugins/viz-platform/README.md) *NEW*
+#### [viz-platform](./plugins/viz-platform/README.md) *NEW in v4.0.0*
 **Dash Mantine Components Validation and Theming**
 
 Visualization toolkit with version-locked component validation and design token theming.
@@ -157,7 +170,7 @@ Comprehensive NetBox REST API integration for infrastructure management.
 | Virtualization | Clusters, VMs, Interfaces |
 | Extras | Tags, Custom Fields, Audit Log |
 
-### Data Platform MCP Server (shared) *NEW*
+### Data Platform MCP Server (shared) *NEW in v4.0.0*
 
 pandas, PostgreSQL/PostGIS, and dbt integration for data engineering.
 
@@ -168,7 +181,7 @@ pandas, PostgreSQL/PostGIS, and dbt integration for data engineering.
 | PostGIS | `st_tables`, `st_geometry_type`, `st_srid`, `st_extent` |
 | dbt | `dbt_parse`, `dbt_run`, `dbt_test`, `dbt_build`, `dbt_compile`, `dbt_ls`, `dbt_docs_generate`, `dbt_lineage` |
 
-### Viz Platform MCP Server (shared) *NEW*
+### Viz Platform MCP Server (shared) *NEW in v4.0.0*
 
 Dash Mantine Components validation and visualization tools.
 
@@ -180,6 +193,16 @@ Dash Mantine Components validation and visualization tools.
 | Theme | `theme_create`, `theme_extend`, `theme_validate`, `theme_export_css`, `theme_list`, `theme_activate` |
 | Page | `page_create`, `page_add_navbar`, `page_set_auth`, `page_list`, `page_get_app_config` |
 
+### Contract Validator MCP Server (shared) *NEW in v5.0.0*
+
+Cross-plugin compatibility validation tools.
+
+| Category | Tools |
+|----------|-------|
+| Parse | `parse_plugin_interface`, `parse_claude_md_agents` |
+| Validation | `validate_compatibility`, `validate_agent_refs`, `validate_data_flow` |
+| Report | `generate_compatibility_report`, `list_issues` |
+
 ## Installation
 
 ### Prerequisites
@@ -278,6 +301,7 @@ After installing plugins, the `/plugin` command may show `(no content)` - this i
 | cmdb-assistant | `/cmdb-assistant:cmdb-search` |
 | data-platform | `/data-platform:ingest` |
 | viz-platform | `/viz-platform:chart` |
+| contract-validator | `/contract-validator:validate-contracts` |
 
 ## Repository Structure
 
@@ -289,14 +313,16 @@ leo-claude-mktplace/
 │   ├── gitea/                     # Gitea MCP (issues, PRs, wiki)
 │   ├── netbox/                    # NetBox MCP (CMDB)
 │   ├── data-platform/             # Data engineering (pandas, PostgreSQL, dbt)
-│   └── viz-platform/              # Visualization (DMC, Plotly, theming)
+│   ├── viz-platform/              # Visualization (DMC, Plotly, theming)
+│   └── contract-validator/        # Cross-plugin validation (v5.0.0)
 ├── plugins/                       # All plugins
 │   ├── projman/                   # Sprint management
 │   ├── git-flow/                  # Git workflow automation
 │   ├── pr-review/                 # PR review
 │   ├── clarity-assist/            # Prompt optimization
 │   ├── data-platform/             # Data engineering
-│   ├── viz-platform/              # Visualization (NEW)
+│   ├── viz-platform/              # Visualization
+│   ├── contract-validator/        # Cross-plugin validation (NEW)
 │   ├── claude-config-maintainer/  # CLAUDE.md optimization
 │   ├── cmdb-assistant/            # NetBox CMDB integration
 │   ├── doc-guardian/              # Documentation drift detection

mcp-servers/contract-validator/run.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CACHE_VENV="$HOME/.cache/claude-mcp-venvs/leo-claude-mktplace/contract-validator/.venv"
+LOCAL_VENV="$SCRIPT_DIR/.venv"
+
+if [[ -f "$CACHE_VENV/bin/python" ]]; then
+    PYTHON="$CACHE_VENV/bin/python"
+elif [[ -f "$LOCAL_VENV/bin/python" ]]; then
+    PYTHON="$LOCAL_VENV/bin/python"
+else
+    echo "ERROR: No venv found. Run: ./scripts/setup-venvs.sh" >&2
+    exit 1
+fi
+
+cd "$SCRIPT_DIR"
+export PYTHONPATH="$SCRIPT_DIR"
+exec "$PYTHON" -m mcp_server.server "$@"

mcp-servers/data-platform/mcp_server/pandas_tools.py

@@ -330,7 +330,7 @@ class PandasTools:
             return {'error': f'DataFrame not found: {data_ref}'}
 
         try:
-            filtered = df.query(condition)
+            filtered = df.query(condition).reset_index(drop=True)
             result_name = name or f"{data_ref}_filtered"
             return self._check_and_store(
                 filtered,

mcp-servers/data-platform/run.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CACHE_VENV="$HOME/.cache/claude-mcp-venvs/leo-claude-mktplace/data-platform/.venv"
+LOCAL_VENV="$SCRIPT_DIR/.venv"
+
+if [[ -f "$CACHE_VENV/bin/python" ]]; then
+    PYTHON="$CACHE_VENV/bin/python"
+elif [[ -f "$LOCAL_VENV/bin/python" ]]; then
+    PYTHON="$LOCAL_VENV/bin/python"
+else
+    echo "ERROR: No venv found. Run: ./scripts/setup-venvs.sh" >&2
+    exit 1
+fi
+
+cd "$SCRIPT_DIR"
+export PYTHONPATH="$SCRIPT_DIR"
+exec "$PYTHON" -m mcp_server.server "$@"

mcp-servers/gitea/mcp_server/gitea_client.py

@@ -787,3 +787,42 @@ class GiteaClient:
         response = self.session.post(url, json=data)
         response.raise_for_status()
         return response.json()
+
+    def create_pull_request(
+        self,
+        title: str,
+        body: str,
+        head: str,
+        base: str,
+        labels: Optional[List[str]] = None,
+        repo: Optional[str] = None
+    ) -> Dict:
+        """
+        Create a new pull request.
+
+        Args:
+            title: PR title
+            body: PR description/body
+            head: Source branch name (the branch with changes)
+            base: Target branch name (the branch to merge into)
+            labels: Optional list of label names
+            repo: Repository in 'owner/repo' format
+
+        Returns:
+            Created pull request dictionary
+        """
+        owner, target_repo = self._parse_repo(repo)
+        url = f"{self.base_url}/repos/{owner}/{target_repo}/pulls"
+        data = {
+            'title': title,
+            'body': body,
+            'head': head,
+            'base': base
+        }
+        if labels:
+            label_ids = self._resolve_label_ids(labels, owner, target_repo)
+            data['labels'] = label_ids
+        logger.info(f"Creating PR '{title}' in {owner}/{target_repo}: {head} -> {base}")
+        response = self.session.post(url, json=data)
+        response.raise_for_status()
+        return response.json()

mcp-servers/gitea/mcp_server/server.py

@@ -844,6 +844,41 @@ class GiteaMCPServer:
                         },
                         "required": ["pr_number", "body"]
                     }
+                ),
+                Tool(
+                    name="create_pull_request",
+                    description="Create a new pull request",
+                    inputSchema={
+                        "type": "object",
+                        "properties": {
+                            "title": {
+                                "type": "string",
+                                "description": "PR title"
+                            },
+                            "body": {
+                                "type": "string",
+                                "description": "PR description/body"
+                            },
+                            "head": {
+                                "type": "string",
+                                "description": "Source branch name (the branch with changes)"
+                            },
+                            "base": {
+                                "type": "string",
+                                "description": "Target branch name (the branch to merge into)"
+                            },
+                            "labels": {
+                                "type": "array",
+                                "items": {"type": "string"},
+                                "description": "Optional list of label names"
+                            },
+                            "repo": {
+                                "type": "string",
+                                "description": "Repository name (owner/repo format)"
+                            }
+                        },
+                        "required": ["title", "body", "head", "base"]
+                    }
                 )
             ]
 
@@ -959,6 +994,8 @@ class GiteaMCPServer:
                     result = await self.pr_tools.create_pr_review(**arguments)
                 elif name == "add_pr_comment":
                     result = await self.pr_tools.add_pr_comment(**arguments)
+                elif name == "create_pull_request":
+                    result = await self.pr_tools.create_pull_request(**arguments)
                 else:
                     raise ValueError(f"Unknown tool: {name}")
 

mcp-servers/gitea/mcp_server/tools/issues.py

@@ -66,7 +66,13 @@ class IssueTools:
             return operation in ['list_issues', 'get_issue', 'get_labels', 'create_issue']
 
         # Development branches (full access)
-        if branch in ['development', 'develop'] or branch.startswith(('feat/', 'feature/', 'dev/')):
+        # Include all common feature/fix branch patterns
+        dev_prefixes = (
+            'feat/', 'feature/', 'dev/',
+            'fix/', 'bugfix/', 'hotfix/',
+            'chore/', 'refactor/', 'docs/', 'test/'
+        )
+        if branch in ['development', 'develop'] or branch.startswith(dev_prefixes):
             return True
 
         # Unknown branch - be restrictive

mcp-servers/gitea/mcp_server/tools/pull_requests.py

@@ -69,7 +69,13 @@ class PullRequestTools:
             return operation in read_ops + ['add_pr_comment']
 
         # Development branches (full access)
-        if branch in ['development', 'develop'] or branch.startswith(('feat/', 'feature/', 'dev/')):
+        # Include all common feature/fix branch patterns
+        dev_prefixes = (
+            'feat/', 'feature/', 'dev/',
+            'fix/', 'bugfix/', 'hotfix/',
+            'chore/', 'refactor/', 'docs/', 'test/'
+        )
+        if branch in ['development', 'develop'] or branch.startswith(dev_prefixes):
             return True
 
         # Unknown branch - be restrictive
@@ -272,3 +278,42 @@ class PullRequestTools:
             None,
             lambda: self.gitea.add_pr_comment(pr_number, body, repo)
         )
+
+    async def create_pull_request(
+        self,
+        title: str,
+        body: str,
+        head: str,
+        base: str,
+        labels: Optional[List[str]] = None,
+        repo: Optional[str] = None
+    ) -> Dict:
+        """
+        Create a new pull request (async wrapper with branch check).
+
+        Args:
+            title: PR title
+            body: PR description/body
+            head: Source branch name (the branch with changes)
+            base: Target branch name (the branch to merge into)
+            labels: Optional list of label names
+            repo: Override configured repo (for PMO multi-repo)
+
+        Returns:
+            Created pull request dictionary
+
+        Raises:
+            PermissionError: If operation not allowed on current branch
+        """
+        if not self._check_branch_permissions('create_pull_request'):
+            branch = self._get_current_branch()
+            raise PermissionError(
+                f"Cannot create PR on branch '{branch}'. "
+                f"Switch to a development or feature branch to create PRs."
+            )
+
+        loop = asyncio.get_event_loop()
+        return await loop.run_in_executor(
+            None,
+            lambda: self.gitea.create_pull_request(title, body, head, base, labels, repo)
+        )

mcp-servers/gitea/run.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CACHE_VENV="$HOME/.cache/claude-mcp-venvs/leo-claude-mktplace/gitea/.venv"
+LOCAL_VENV="$SCRIPT_DIR/.venv"
+
+if [[ -f "$CACHE_VENV/bin/python" ]]; then
+    PYTHON="$CACHE_VENV/bin/python"
+elif [[ -f "$LOCAL_VENV/bin/python" ]]; then
+    PYTHON="$LOCAL_VENV/bin/python"
+else
+    echo "ERROR: No venv found. Run: ./scripts/setup-venvs.sh" >&2
+    exit 1
+fi
+
+cd "$SCRIPT_DIR"
+export PYTHONPATH="$SCRIPT_DIR"
+exec "$PYTHON" -m mcp_server.server "$@"

mcp-servers/netbox/run.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CACHE_VENV="$HOME/.cache/claude-mcp-venvs/leo-claude-mktplace/netbox/.venv"
+LOCAL_VENV="$SCRIPT_DIR/.venv"
+
+if [[ -f "$CACHE_VENV/bin/python" ]]; then
+    PYTHON="$CACHE_VENV/bin/python"
+elif [[ -f "$LOCAL_VENV/bin/python" ]]; then
+    PYTHON="$LOCAL_VENV/bin/python"
+else
+    echo "ERROR: No venv found. Run: ./scripts/setup-venvs.sh" >&2
+    exit 1
+fi
+
+cd "$SCRIPT_DIR"
+export PYTHONPATH="$SCRIPT_DIR"
+exec "$PYTHON" -m mcp_server.server "$@"

mcp-servers/viz-platform/run.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+CACHE_VENV="$HOME/.cache/claude-mcp-venvs/leo-claude-mktplace/viz-platform/.venv"
+LOCAL_VENV="$SCRIPT_DIR/.venv"
+
+if [[ -f "$CACHE_VENV/bin/python" ]]; then
+    PYTHON="$CACHE_VENV/bin/python"
+elif [[ -f "$LOCAL_VENV/bin/python" ]]; then
+    PYTHON="$LOCAL_VENV/bin/python"
+else
+    echo "ERROR: No venv found. Run: ./scripts/setup-venvs.sh" >&2
+    exit 1
+fi
+
+cd "$SCRIPT_DIR"
+export PYTHONPATH="$SCRIPT_DIR"
+exec "$PYTHON" -m mcp_server.server "$@"

plugins/cmdb-assistant/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "cmdb-assistant",
-  "version": "1.0.0",
-  "description": "NetBox CMDB integration for infrastructure management - query, create, update, and manage network devices, IP addresses, sites, and more",
+  "version": "1.1.0",
+  "description": "NetBox CMDB integration with data quality validation - query, create, update, and manage network devices, IP addresses, sites, and more with best practices enforcement",
   "author": {
     "name": "Leo Miranda",
     "email": "leobmiranda@gmail.com"
@@ -15,7 +15,9 @@
     "infrastructure",
     "network",
     "ipam",
-    "dcim"
+    "dcim",
+    "data-quality",
+    "validation"
   ],
   "commands": ["./commands/"],
   "mcpServers": ["./.mcp.json"]

plugins/cmdb-assistant/.mcp.json

@@ -1,12 +1,8 @@
 {
   "mcpServers": {
     "netbox": {
-      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/netbox/.venv/bin/python",
-      "args": ["-m", "mcp_server.server"],
-      "cwd": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/netbox",
-      "env": {
-        "PYTHONPATH": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/netbox"
-      }
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/netbox/run.sh",
+      "args": []
     }
   }
 }

plugins/cmdb-assistant/README.md

@@ -2,6 +2,14 @@
 
 A Claude Code plugin for NetBox CMDB integration - query, create, update, and manage your network infrastructure directly from Claude Code.
 
+## What's New in v1.1.0
+
+- **Data Quality Validation**: Hooks for SessionStart and PreToolUse that check data quality and warn about missing fields
+- **Best Practices Skill**: Reference documentation for NetBox patterns (naming conventions, dependency order, role management)
+- **`/cmdb-audit`**: Analyze data quality across VMs, devices, naming conventions, and roles
+- **`/cmdb-register`**: Register the current machine into NetBox with all running applications (Docker containers, systemd services)
+- **`/cmdb-sync`**: Synchronize existing machine state with NetBox (detect drift, update with confirmation)
+
 ## Features
 
 - **Full CRUD Operations**: Create, read, update, and delete across all NetBox modules
@@ -9,6 +17,9 @@ A Claude Code plugin for NetBox CMDB integration - query, create, update, and ma
 - **IP Management**: Allocate IPs, manage prefixes, track VLANs
 - **Infrastructure Documentation**: Document servers, network devices, and connections
 - **Audit Trail**: Review changes and maintain infrastructure history
+- **Data Quality Validation**: Proactive checks for missing site, tenant, platform assignments
+- **Machine Registration**: Auto-discover and register servers with running applications
+- **Drift Detection**: Sync machine state and detect changes over time
 
 ## Installation
 
@@ -40,10 +51,14 @@ Add to your Claude Code plugins or marketplace configuration.
 
 | Command | Description |
 |---------|-------------|
+| `/initial-setup` | Interactive setup wizard for NetBox MCP server |
 | `/cmdb-search <query>` | Search for devices, IPs, sites, or any CMDB object |
 | `/cmdb-device <action>` | Manage network devices (list, create, update, delete) |
 | `/cmdb-ip <action>` | Manage IP addresses and prefixes |
 | `/cmdb-site <action>` | Manage sites and locations |
+| `/cmdb-audit [scope]` | Data quality analysis (all, vms, devices, naming, roles) |
+| `/cmdb-register` | Register current machine into NetBox with running apps |
+| `/cmdb-sync` | Sync machine state with NetBox (detect drift, update) |
 
 ## Agent
 
@@ -103,6 +118,15 @@ This plugin provides access to the full NetBox API:
 - **Wireless**: WLANs, Wireless Links
 - **Extras**: Tags, Custom Fields, Journal Entries, Audit Log
 
+## Hooks
+
+| Event | Purpose |
+|-------|---------|
+| `SessionStart` | Test NetBox connectivity, report data quality issues |
+| `PreToolUse` | Validate VM/device parameters before create/update |
+
+Hooks are **non-blocking** - they emit warnings but never prevent operations.
+
 ## Architecture
 
 ```
@@ -115,13 +139,23 @@ cmdb-assistant/
 │   ├── cmdb-search.md       # Search command
 │   ├── cmdb-device.md       # Device management
 │   ├── cmdb-ip.md           # IP management
-│   └── cmdb-site.md         # Site management
+│   ├── cmdb-site.md         # Site management
+│   ├── cmdb-audit.md        # Data quality audit (NEW)
+│   ├── cmdb-register.md     # Machine registration (NEW)
+│   └── cmdb-sync.md         # Machine sync (NEW)
+├── hooks/
+│   ├── hooks.json           # Hook configuration
+│   ├── startup-check.sh     # SessionStart validation
+│   └── validate-input.sh    # PreToolUse validation
+├── skills/
+│   └── netbox-patterns/
+│       └── SKILL.md         # NetBox best practices reference
 ├── agents/
 │   └── cmdb-assistant.md    # Main assistant agent
 └── README.md
 ```
 
-The plugin uses the shared NetBox MCP server at `../mcp-servers/netbox/`.
+The plugin uses the shared NetBox MCP server at `mcp-servers/netbox/`.
 
 ## Configuration
 

plugins/cmdb-assistant/agents/cmdb-assistant.md

@@ -76,3 +76,132 @@ When presenting data:
 - Suggest corrective actions
 - For permission errors, note what access is needed
 - For validation errors, explain required fields/formats
+
+## Data Quality Validation
+
+**IMPORTANT:** Load the `netbox-patterns` skill for best practice reference.
+
+Before ANY create or update operation, validate against NetBox best practices:
+
+### VM Operations
+
+**Required checks before `virt_create_vm` or `virt_update_vm`:**
+
+1. **Cluster/Site Assignment** - VMs must have either cluster or site
+2. **Tenant Assignment** - Recommend if not provided
+3. **Platform Assignment** - Recommend for OS tracking
+4. **Naming Convention** - Check against `{env}-{app}-{number}` pattern
+5. **Role Assignment** - Recommend appropriate role
+
+**If user provides no site/tenant, ASK:**
+
+> "This VM has no site or tenant assigned. NetBox best practices recommend:
+> - **Site**: For location-based queries and power budgeting
+> - **Tenant**: For resource isolation and ownership tracking
+>
+> Would you like me to:
+> 1. Assign to an existing site/tenant (list available)
+> 2. Create new site/tenant first
+> 3. Proceed without (not recommended for production use)"
+
+### Device Operations
+
+**Required checks before `dcim_create_device` or `dcim_update_device`:**
+
+1. **Site is REQUIRED** - Fail without it
+2. **Platform Assignment** - Recommend for OS tracking
+3. **Naming Convention** - Check against `{role}-{location}-{number}` pattern
+4. **Role Assignment** - Ensure appropriate role selected
+5. **After Creation** - Offer to set primary IP
+
+### Cluster Operations
+
+**Required checks before `virt_create_cluster`:**
+
+1. **Site Scope** - Recommend assigning to site
+2. **Cluster Type** - Ensure appropriate type selected
+3. **Device Association** - Recommend linking to host device
+
+### Role Management
+
+**Before creating a new device role:**
+
+1. List existing roles with `dcim_list_device_roles`
+2. Check if a more general role already exists
+3. Recommend role consolidation if >10 specific roles exist
+
+**Example guidance:**
+
+> "You're creating role 'nginx-web-server'. An existing 'web-server' role exists.
+> Consider using 'web-server' and tracking nginx via the platform field instead.
+> This reduces role fragmentation and improves maintainability."
+
+## Dependency Order Enforcement
+
+When creating multiple objects, follow this order:
+
+```
+1. Regions → Sites → Locations → Racks
+2. Tenant Groups → Tenants
+3. Manufacturers → Device Types
+4. Device Roles, Platforms
+5. Devices (with site, role, type)
+6. Clusters (with type, optional site)
+7. VMs (with cluster)
+8. Interfaces → IP Addresses → Primary IP assignment
+```
+
+**CRITICAL Rules:**
+- NEVER create a VM before its cluster exists
+- NEVER create a device before its site exists
+- NEVER create an interface before its device exists
+- NEVER create an IP before its interface exists (if assigning)
+
+## Naming Convention Enforcement
+
+When user provides a name, check against patterns:
+
+| Object Type | Pattern | Example |
+|-------------|---------|---------|
+| Device | `{role}-{site}-{number}` | `web-dc1-01` |
+| VM | `{env}-{app}-{number}` or `{prefix}_{service}` | `prod-api-01` |
+| Cluster | `{site}-{type}` | `dc1-vmware`, `home-docker` |
+| Prefix | Include purpose in description | "Production /24 for web tier" |
+
+**If name doesn't match patterns, warn:**
+
+> "The name 'HotServ' doesn't follow naming conventions.
+> Suggested: `prod-hotserv-01` or `hotserv-cloud-01`.
+> Consistent naming improves searchability and automation compatibility.
+> Proceed with original name? [Y/n]"
+
+## Duplicate Prevention
+
+Before creating objects, always check for existing duplicates:
+
+```
+# Before creating device
+dcim_list_devices name=<proposed-name>
+
+# Before creating VM
+virt_list_vms name=<proposed-name>
+
+# Before creating prefix
+ipam_list_prefixes prefix=<proposed-prefix>
+```
+
+If duplicate found, inform user and suggest update instead of create.
+
+## Available Commands
+
+Users can invoke these commands for structured workflows:
+
+| Command | Purpose |
+|---------|---------|
+| `/cmdb-search <query>` | Search across all CMDB objects |
+| `/cmdb-device <action>` | Device CRUD operations |
+| `/cmdb-ip <action>` | IP address and prefix management |
+| `/cmdb-site <action>` | Site and location management |
+| `/cmdb-audit [scope]` | Data quality analysis |
+| `/cmdb-register` | Register current machine |
+| `/cmdb-sync` | Sync machine state with NetBox |

plugins/cmdb-assistant/commands/cmdb-audit.md

@@ -0,0 +1,195 @@
+---
+description: Audit NetBox data quality and identify consistency issues
+---
+
+# CMDB Data Quality Audit
+
+Analyze NetBox data for quality issues and best practice violations.
+
+## Usage
+
+```
+/cmdb-audit [scope]
+```
+
+**Scopes:**
+- `all` (default) - Full audit across all categories
+- `vms` - Virtual machines only
+- `devices` - Physical devices only
+- `naming` - Naming convention analysis
+- `roles` - Role fragmentation analysis
+
+## Instructions
+
+You are a data quality auditor for NetBox. Your job is to identify consistency issues and best practice violations.
+
+**IMPORTANT:** Load the `netbox-patterns` skill for best practice reference.
+
+### Phase 1: Data Collection
+
+Run these MCP tool calls to gather data for analysis:
+
+```
+1. virt_list_vms (no filters - get all)
+2. dcim_list_devices (no filters - get all)
+3. virt_list_clusters (no filters)
+4. dcim_list_sites
+5. tenancy_list_tenants
+6. dcim_list_device_roles
+7. dcim_list_platforms
+```
+
+Store the results for analysis.
+
+### Phase 2: Quality Checks
+
+Analyze collected data for these issues by severity:
+
+#### CRITICAL Issues (must fix immediately)
+
+| Check | Detection |
+|-------|-----------|
+| VMs without cluster | `cluster` field is null AND `site` field is null |
+| Devices without site | `site` field is null |
+| Active devices without primary IP | `status=active` AND `primary_ip4` is null AND `primary_ip6` is null |
+
+#### HIGH Issues (should fix soon)
+
+| Check | Detection |
+|-------|-----------|
+| VMs without site | VM has no site (neither direct nor via cluster.site) |
+| VMs without tenant | `tenant` field is null |
+| Devices without platform | `platform` field is null |
+| Clusters not scoped to site | `site` field is null on cluster |
+| VMs without role | `role` field is null |
+
+#### MEDIUM Issues (plan to address)
+
+| Check | Detection |
+|-------|-----------|
+| Inconsistent naming | Names don't match patterns: devices=`{role}-{site}-{num}`, VMs=`{env}-{app}-{num}` |
+| Role fragmentation | More than 10 device roles with <3 assignments each |
+| Missing tags on production | Active resources without any tags |
+| Mixed naming separators | Some names use `_`, others use `-` |
+
+#### LOW Issues (informational)
+
+| Check | Detection |
+|-------|-----------|
+| Docker containers as VMs | Cluster type is "Docker Compose" - document this modeling choice |
+| VMs without description | `description` field is empty |
+| Sites without physical address | `physical_address` is empty |
+| Devices without serial | `serial` field is empty |
+
+### Phase 3: Naming Convention Analysis
+
+For naming scope, analyze patterns:
+
+1. **Extract naming patterns** from existing objects
+2. **Identify dominant patterns** (most common conventions)
+3. **Flag outliers** that don't match dominant patterns
+4. **Suggest standardization** based on best practices
+
+**Expected Patterns:**
+- Devices: `{role}-{location}-{number}` (e.g., `web-dc1-01`)
+- VMs: `{prefix}_{service}` or `{env}-{app}-{number}` (e.g., `prod-api-01`)
+- Clusters: `{site}-{type}` (e.g., `home-docker`)
+
+### Phase 4: Role Analysis
+
+For roles scope, analyze fragmentation:
+
+1. **List all device roles** with assignment counts
+2. **Identify single-use roles** (only 1 device/VM)
+3. **Identify similar roles** that could be consolidated
+4. **Suggest consolidation** based on patterns
+
+**Red Flags:**
+- More than 15 highly specific roles
+- Roles with technology in name (use platform instead)
+- Roles that duplicate functionality
+
+### Phase 5: Report Generation
+
+Present findings in this structure:
+
+```markdown
+## CMDB Data Quality Audit Report
+
+**Generated:** [timestamp]
+**Scope:** [scope parameter]
+
+### Summary
+
+| Metric | Count |
+|--------|-------|
+| Total VMs | X |
+| Total Devices | Y |
+| Total Clusters | Z |
+| **Total Issues** | **N** |
+
+| Severity | Count |
+|----------|-------|
+| Critical | A |
+| High | B |
+| Medium | C |
+| Low | D |
+
+### Critical Issues
+
+[List each with specific object names and IDs]
+
+**Example:**
+- VM `HotServ` (ID: 1) - No cluster or site assignment
+- Device `server-01` (ID: 5) - No site assignment
+
+### High Issues
+
+[List each with specific object names]
+
+### Medium Issues
+
+[Grouped by category with counts]
+
+### Recommendations
+
+1. **[Most impactful fix]** - affects N objects
+2. **[Second priority]** - affects M objects
+...
+
+### Quick Fixes
+
+Commands to fix common issues:
+
+```
+# Assign site to VM
+virt_update_vm id=X site=Y
+
+# Assign platform to device
+dcim_update_device id=X platform=Y
+```
+
+### Next Steps
+
+- Run `/cmdb-register` to properly register new machines
+- Use `/cmdb-sync` to update existing registrations
+- Consider bulk updates via NetBox web UI for >10 items
+```
+
+## Scope-Specific Instructions
+
+### For `vms` scope:
+Focus only on Virtual Machine checks. Skip device and role analysis.
+
+### For `devices` scope:
+Focus only on Device checks. Skip VM and cluster analysis.
+
+### For `naming` scope:
+Focus on naming convention analysis across all objects. Generate detailed pattern report.
+
+### For `roles` scope:
+Focus on role fragmentation analysis. Generate consolidation recommendations.
+
+## User Request
+
+$ARGUMENTS

plugins/cmdb-assistant/commands/cmdb-register.md

@@ -0,0 +1,322 @@
+---
+description: Register the current machine into NetBox with all running applications
+---
+
+# CMDB Machine Registration
+
+Register the current machine into NetBox, including hardware info, network interfaces, and running applications (Docker containers, services).
+
+## Usage
+
+```
+/cmdb-register [--site <site-name>] [--tenant <tenant-name>] [--role <role-name>]
+```
+
+**Options:**
+- `--site <name>`: Site to assign (will prompt if not provided)
+- `--tenant <name>`: Tenant for resource isolation (optional)
+- `--role <name>`: Device role (default: auto-detect based on services)
+
+## Instructions
+
+You are registering the current machine into NetBox. This is a multi-phase process that discovers local system information and creates corresponding NetBox objects.
+
+**IMPORTANT:** Load the `netbox-patterns` skill for best practice reference.
+
+### Phase 1: System Discovery (via Bash)
+
+Gather system information using these commands:
+
+#### 1.1 Basic Device Info
+
+```bash
+# Hostname
+hostname
+
+# OS/Platform info
+cat /etc/os-release 2>/dev/null || uname -a
+
+# Hardware model (varies by system)
+# Raspberry Pi:
+cat /proc/device-tree/model 2>/dev/null || echo "Unknown"
+
+# x86 systems:
+cat /sys/class/dmi/id/product_name 2>/dev/null || echo "Unknown"
+
+# Serial number
+# Raspberry Pi:
+cat /proc/device-tree/serial-number 2>/dev/null || cat /proc/cpuinfo | grep Serial | cut -d: -f2 | tr -d ' ' 2>/dev/null
+
+# x86 systems:
+cat /sys/class/dmi/id/product_serial 2>/dev/null || echo "Unknown"
+
+# CPU info
+nproc
+
+# Memory (MB)
+free -m | awk '/Mem:/ {print $2}'
+
+# Disk (GB, root filesystem)
+df -BG / | awk 'NR==2 {print $2}' | tr -d 'G'
+```
+
+#### 1.2 Network Interfaces
+
+```bash
+# Get interfaces with IPs (JSON format)
+ip -j addr show 2>/dev/null || ip addr show
+
+# Get default gateway interface
+ip route | grep default | awk '{print $5}' | head -1
+
+# Get MAC addresses
+ip -j link show 2>/dev/null || ip link show
+```
+
+#### 1.3 Running Applications
+
+```bash
+# Docker containers (if docker available)
+docker ps --format '{"name":"{{.Names}}","image":"{{.Image}}","status":"{{.Status}}","ports":"{{.Ports}}"}' 2>/dev/null || echo "Docker not available"
+
+# Docker Compose projects (check common locations)
+find ~/apps /home/*/apps -name "docker-compose.yml" -o -name "docker-compose.yaml" 2>/dev/null | head -20
+
+# Systemd services (running)
+systemctl list-units --type=service --state=running --no-pager --plain 2>/dev/null | grep -v "^UNIT" | head -30
+```
+
+### Phase 2: Pre-Registration Checks (via MCP)
+
+Before creating objects, verify prerequisites:
+
+#### 2.1 Check if Device Already Exists
+
+```
+dcim_list_devices name=<hostname>
+```
+
+**If device exists:**
+- Inform user and suggest `/cmdb-sync` instead
+- Ask if they want to proceed with re-registration (will update existing)
+
+#### 2.2 Verify/Create Site
+
+If `--site` provided:
+```
+dcim_list_sites name=<site-name>
+```
+
+If site doesn't exist, ask user if they want to create it.
+
+If no site provided, list available sites and ask user to choose:
+```
+dcim_list_sites
+```
+
+#### 2.3 Verify/Create Platform
+
+Based on OS detected, check if platform exists:
+```
+dcim_list_platforms name=<platform-name>
+```
+
+**Platform naming:**
+- `Raspberry Pi OS (Bookworm)` for Raspberry Pi
+- `Ubuntu 24.04 LTS` for Ubuntu
+- `Debian 12` for Debian
+- Use format: `{OS Name} {Version}`
+
+If platform doesn't exist, create it:
+```
+dcim_create_platform name=<platform-name> slug=<slug>
+```
+
+#### 2.4 Verify/Create Device Role
+
+Based on detected services:
+- If Docker containers found → `Docker Host`
+- If only basic services → `Server`
+- If specific role specified → Use that
+
+```
+dcim_list_device_roles name=<role-name>
+```
+
+### Phase 3: Device Registration (via MCP)
+
+#### 3.1 Get/Create Manufacturer and Device Type
+
+For Raspberry Pi:
+```
+dcim_list_manufacturers name="Raspberry Pi Foundation"
+dcim_list_device_types manufacturer_id=X model="Raspberry Pi 4 Model B"
+```
+
+Create if not exists.
+
+For generic x86:
+```
+dcim_list_manufacturers name=<detected-manufacturer>
+```
+
+#### 3.2 Create Device
+
+```
+dcim_create_device
+  name=<hostname>
+  device_type=<device_type_id>
+  role=<role_id>
+  site=<site_id>
+  platform=<platform_id>
+  tenant=<tenant_id>  # if provided
+  serial=<serial>
+  description="Registered via cmdb-assistant"
+```
+
+#### 3.3 Create Interfaces
+
+For each network interface discovered:
+```
+dcim_create_interface
+  device=<device_id>
+  name=<interface_name>  # eth0, wlan0, tailscale0, etc.
+  type=<type>            # 1000base-t, virtual, other
+  mac_address=<mac>
+  enabled=true
+```
+
+**Interface type mapping:**
+- `eth*`, `enp*` → `1000base-t`
+- `wlan*` → `ieee802.11ax` (or appropriate wifi type)
+- `tailscale*`, `docker*`, `br-*` → `virtual`
+- `lo` → skip (loopback)
+
+#### 3.4 Create IP Addresses
+
+For each IP on each interface:
+```
+ipam_create_ip_address
+  address=<ip/prefix>    # e.g., "192.168.1.100/24"
+  assigned_object_type="dcim.interface"
+  assigned_object_id=<interface_id>
+  status="active"
+  description="Discovered via cmdb-register"
+```
+
+#### 3.5 Set Primary IP
+
+Identify primary IP (interface with default route):
+```
+dcim_update_device
+  id=<device_id>
+  primary_ip4=<primary_ip_id>
+```
+
+### Phase 4: Container Registration (via MCP)
+
+If Docker containers were discovered:
+
+#### 4.1 Create/Get Cluster Type
+
+```
+virt_list_cluster_types name="Docker Compose"
+```
+
+Create if not exists:
+```
+virt_create_cluster_type name="Docker Compose" slug="docker-compose"
+```
+
+#### 4.2 Create Cluster
+
+For each Docker Compose project directory found:
+```
+virt_create_cluster
+  name=<project-name>  # e.g., "apps-hotport"
+  type=<cluster_type_id>
+  site=<site_id>
+  description="Docker Compose stack on <hostname>"
+```
+
+#### 4.3 Create VMs for Containers
+
+For each running container:
+```
+virt_create_vm
+  name=<container_name>
+  cluster=<cluster_id>
+  site=<site_id>
+  role=<role_id>        # Map container function to role
+  status="active"
+  vcpus=<cpu_shares>    # Default 1.0 if unknown
+  memory=<memory_mb>    # Default 256 if unknown
+  disk=<disk_gb>        # Default 5 if unknown
+  description=<container purpose>
+  comments=<image, ports, volumes info>
+```
+
+**Container role mapping:**
+- `*caddy*`, `*nginx*`, `*traefik*` → "Reverse Proxy"
+- `*db*`, `*postgres*`, `*mysql*`, `*redis*` → "Database"
+- `*webui*`, `*frontend*` → "Web Application"
+- Others → Infer from image name or use generic "Container"
+
+### Phase 5: Documentation
+
+#### 5.1 Add Journal Entry
+
+```
+extras_create_journal_entry
+  assigned_object_type="dcim.device"
+  assigned_object_id=<device_id>
+  comments="Device registered via /cmdb-register command\n\nDiscovered:\n- X network interfaces\n- Y IP addresses\n- Z Docker containers"
+```
+
+### Phase 6: Summary Report
+
+Present registration summary:
+
+```markdown
+## Machine Registration Complete
+
+### Device Created
+- **Name:** <hostname>
+- **Site:** <site>
+- **Platform:** <platform>
+- **Role:** <role>
+- **ID:** <device_id>
+- **URL:** https://netbox.example.com/dcim/devices/<id>/
+
+### Network Interfaces
+| Interface | Type | MAC | IP Address |
+|-----------|------|-----|------------|
+| eth0 | 1000base-t | aa:bb:cc:dd:ee:ff | 192.168.1.100/24 |
+| tailscale0 | virtual | - | 100.x.x.x/32 |
+
+### Primary IP: 192.168.1.100
+
+### Docker Containers Registered (if applicable)
+**Cluster:** <cluster_name> (ID: <cluster_id>)
+
+| Container | Role | vCPUs | Memory | Status |
+|-----------|------|-------|--------|--------|
+| media_jellyfin | Media Server | 2.0 | 2048MB | Active |
+| media_sonarr | Media Management | 1.0 | 512MB | Active |
+
+### Next Steps
+- Run `/cmdb-sync` periodically to keep data current
+- Run `/cmdb-audit` to check data quality
+- Add tags for classification (env:*, team:*, etc.)
+```
+
+## Error Handling
+
+- **Device already exists:** Suggest `/cmdb-sync` or ask to proceed
+- **Site not found:** List available sites, offer to create new
+- **Docker not available:** Skip container registration, note in summary
+- **Permission denied:** Note which operations failed, suggest fixes
+
+## User Request
+
+$ARGUMENTS

plugins/cmdb-assistant/commands/cmdb-sync.md

@@ -0,0 +1,336 @@
+---
+description: Synchronize current machine state with existing NetBox record
+---
+
+# CMDB Machine Sync
+
+Update an existing NetBox device record with the current machine state. Compares local system information with NetBox and applies changes.
+
+## Usage
+
+```
+/cmdb-sync [--full] [--dry-run]
+```
+
+**Options:**
+- `--full`: Force refresh all fields, even unchanged ones
+- `--dry-run`: Show what would change without applying updates
+
+## Instructions
+
+You are synchronizing the current machine's state with its NetBox record. This involves comparing current system state with stored data and updating differences.
+
+**IMPORTANT:** Load the `netbox-patterns` skill for best practice reference.
+
+### Phase 1: Device Lookup (via MCP)
+
+First, find the existing device record:
+
+```bash
+# Get current hostname
+hostname
+```
+
+```
+dcim_list_devices name=<hostname>
+```
+
+**If device not found:**
+- Inform user: "Device '<hostname>' not found in NetBox"
+- Suggest: "Run `/cmdb-register` to register this machine first"
+- Exit sync
+
+**If device found:**
+- Store device ID and all current field values
+- Fetch interfaces: `dcim_list_interfaces device_id=<device_id>`
+- Fetch IPs: `ipam_list_ip_addresses device_id=<device_id>`
+
+Also check for associated clusters/VMs:
+```
+virt_list_clusters  # Look for cluster associated with this device
+virt_list_vms cluster=<cluster_id>  # If cluster found
+```
+
+### Phase 2: Current State Discovery (via Bash)
+
+Gather current system information (same as `/cmdb-register`):
+
+```bash
+# Device info
+hostname
+cat /etc/os-release 2>/dev/null || uname -a
+nproc
+free -m | awk '/Mem:/ {print $2}'
+df -BG / | awk 'NR==2 {print $2}' | tr -d 'G'
+
+# Network interfaces with IPs
+ip -j addr show 2>/dev/null || ip addr show
+
+# Docker containers
+docker ps --format '{"name":"{{.Names}}","image":"{{.Image}}","status":"{{.Status}}"}' 2>/dev/null || echo "[]"
+```
+
+### Phase 3: Comparison
+
+Compare discovered state with NetBox record:
+
+#### 3.1 Device Attributes
+
+| Field | Compare |
+|-------|---------|
+| Platform | OS version changed? |
+| Status | Still active? |
+| Serial | Match? |
+| Description | Keep existing |
+
+#### 3.2 Network Interfaces
+
+| Change Type | Detection |
+|-------------|-----------|
+| New interface | Interface exists locally but not in NetBox |
+| Removed interface | Interface in NetBox but not locally |
+| Changed MAC | MAC address different |
+| Interface type | Type mismatch |
+
+#### 3.3 IP Addresses
+
+| Change Type | Detection |
+|-------------|-----------|
+| New IP | IP exists locally but not in NetBox |
+| Removed IP | IP in NetBox but not locally (on this device) |
+| Primary IP changed | Default route interface changed |
+
+#### 3.4 Docker Containers
+
+| Change Type | Detection |
+|-------------|-----------|
+| New container | Container running locally but no VM in cluster |
+| Stopped container | VM exists but container not running |
+| Resource change | vCPUs/memory different (if trackable) |
+
+### Phase 4: Diff Report
+
+Present changes to user:
+
+```markdown
+## Sync Diff Report
+
+**Device:** <hostname> (ID: <device_id>)
+**NetBox URL:** https://netbox.example.com/dcim/devices/<id>/
+
+### Device Attributes
+| Field | NetBox Value | Current Value | Action |
+|-------|--------------|---------------|--------|
+| Platform | Ubuntu 22.04 | Ubuntu 24.04 | UPDATE |
+| Status | active | active | - |
+
+### Network Interfaces
+
+#### New Interfaces (will create)
+| Interface | Type | MAC | IPs |
+|-----------|------|-----|-----|
+| tailscale0 | virtual | - | 100.x.x.x/32 |
+
+#### Removed Interfaces (will mark offline)
+| Interface | Type | Reason |
+|-----------|------|--------|
+| eth1 | 1000base-t | Not found locally |
+
+#### Changed Interfaces
+| Interface | Field | Old | New |
+|-----------|-------|-----|-----|
+| eth0 | mac_address | aa:bb:cc:00:00:00 | aa:bb:cc:11:11:11 |
+
+### IP Addresses
+
+#### New IPs (will create)
+- 192.168.1.150/24 on eth0
+
+#### Removed IPs (will unassign)
+- 192.168.1.100/24 from eth0
+
+### Docker Containers
+
+#### New Containers (will create VMs)
+| Container | Image | Role |
+|-----------|-------|------|
+| media_lidarr | linuxserver/lidarr | Media Management |
+
+#### Stopped Containers (will mark offline)
+| Container | Last Status |
+|-----------|-------------|
+| media_bazarr | Exited |
+
+### Summary
+- **Updates:** X
+- **Creates:** Y
+- **Removals/Offline:** Z
+```
+
+### Phase 5: User Confirmation
+
+If not `--dry-run`:
+
+```
+The following changes will be applied:
+- Update device platform to "Ubuntu 24.04"
+- Create interface "tailscale0"
+- Create IP "100.x.x.x/32" on tailscale0
+- Create VM "media_lidarr" in cluster
+- Mark VM "media_bazarr" as offline
+
+Proceed with sync? [Y/n]
+```
+
+**Use AskUserQuestion** to get confirmation.
+
+### Phase 6: Apply Updates (via MCP)
+
+Only if user confirms (or `--full` specified):
+
+#### 6.1 Device Updates
+
+```
+dcim_update_device
+  id=<device_id>
+  platform=<new_platform_id>
+  # ... other changed fields
+```
+
+#### 6.2 Interface Updates
+
+**For new interfaces:**
+```
+dcim_create_interface
+  device=<device_id>
+  name=<interface_name>
+  type=<type>
+  mac_address=<mac>
+  enabled=true
+```
+
+**For removed interfaces:**
+```
+dcim_update_interface
+  id=<interface_id>
+  enabled=false
+  description="Marked offline by cmdb-sync - interface no longer present"
+```
+
+**For changed interfaces:**
+```
+dcim_update_interface
+  id=<interface_id>
+  mac_address=<new_mac>
+```
+
+#### 6.3 IP Address Updates
+
+**For new IPs:**
+```
+ipam_create_ip_address
+  address=<ip/prefix>
+  assigned_object_type="dcim.interface"
+  assigned_object_id=<interface_id>
+  status="active"
+```
+
+**For removed IPs:**
+```
+ipam_update_ip_address
+  id=<ip_id>
+  assigned_object_type=null
+  assigned_object_id=null
+  description="Unassigned by cmdb-sync"
+```
+
+#### 6.4 Primary IP Update
+
+If primary IP changed:
+```
+dcim_update_device
+  id=<device_id>
+  primary_ip4=<new_primary_ip_id>
+```
+
+#### 6.5 Container/VM Updates
+
+**For new containers:**
+```
+virt_create_vm
+  name=<container_name>
+  cluster=<cluster_id>
+  status="active"
+  # ... other fields
+```
+
+**For stopped containers:**
+```
+virt_update_vm
+  id=<vm_id>
+  status="offline"
+  description="Container stopped - detected by cmdb-sync"
+```
+
+### Phase 7: Journal Entry
+
+Document the sync:
+
+```
+extras_create_journal_entry
+  assigned_object_type="dcim.device"
+  assigned_object_id=<device_id>
+  comments="Device synced via /cmdb-sync command\n\nChanges applied:\n- <list of changes>"
+```
+
+### Phase 8: Summary Report
+
+```markdown
+## Sync Complete
+
+**Device:** <hostname>
+**Sync Time:** <timestamp>
+
+### Changes Applied
+- Updated platform: Ubuntu 22.04 → Ubuntu 24.04
+- Created interface: tailscale0 (ID: X)
+- Created IP: 100.x.x.x/32 (ID: Y)
+- Created VM: media_lidarr (ID: Z)
+- Marked VM offline: media_bazarr (ID: W)
+
+### Current State
+- **Interfaces:** 4 (3 active, 1 offline)
+- **IP Addresses:** 5
+- **Containers/VMs:** 8 (7 active, 1 offline)
+
+### Next Sync
+Run `/cmdb-sync` again after:
+- Adding/removing Docker containers
+- Changing network configuration
+- OS upgrades
+```
+
+## Dry Run Mode
+
+If `--dry-run` specified:
+- Complete Phase 1-4 (lookup, discovery, compare, diff report)
+- Skip Phase 5-8 (no confirmation, no updates, no journal)
+- End with: "Dry run complete. No changes applied. Run without --dry-run to apply."
+
+## Full Sync Mode
+
+If `--full` specified:
+- Skip user confirmation
+- Update all fields even if unchanged (force refresh)
+- Useful for ensuring NetBox matches current state exactly
+
+## Error Handling
+
+- **Device not found:** Suggest `/cmdb-register`
+- **Permission denied on updates:** Note which failed, continue with others
+- **Cluster not found:** Offer to create or skip container sync
+- **API errors:** Log error, continue with remaining updates
+
+## User Request
+
+$ARGUMENTS

plugins/cmdb-assistant/hooks/hooks.json

@@ -0,0 +1,21 @@
+{
+  "hooks": {
+    "SessionStart": [
+      {
+        "type": "command",
+        "command": "${CLAUDE_PLUGIN_ROOT}/hooks/startup-check.sh"
+      }
+    ],
+    "PreToolUse": [
+      {
+        "matcher": "mcp__plugin_cmdb-assistant_netbox__virt_create|mcp__plugin_cmdb-assistant_netbox__virt_update|mcp__plugin_cmdb-assistant_netbox__dcim_create|mcp__plugin_cmdb-assistant_netbox__dcim_update",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "${CLAUDE_PLUGIN_ROOT}/hooks/validate-input.sh"
+          }
+        ]
+      }
+    ]
+  }
+}

plugins/cmdb-assistant/hooks/startup-check.sh

@@ -0,0 +1,66 @@
+#!/bin/bash
+# cmdb-assistant SessionStart hook
+# Tests NetBox API connectivity and checks for data quality issues
+# All output MUST have [cmdb-assistant] prefix
+# Non-blocking: always exits 0
+
+set -euo pipefail
+
+PREFIX="[cmdb-assistant]"
+
+# Load NetBox configuration
+NETBOX_CONFIG="$HOME/.config/claude/netbox.env"
+
+if [[ ! -f "$NETBOX_CONFIG" ]]; then
+    echo "$PREFIX NetBox not configured - run /cmdb-assistant:initial-setup"
+    exit 0
+fi
+
+# Source config
+source "$NETBOX_CONFIG"
+
+# Validate required variables
+if [[ -z "${NETBOX_API_URL:-}" ]] || [[ -z "${NETBOX_API_TOKEN:-}" ]]; then
+    echo "$PREFIX Missing NETBOX_API_URL or NETBOX_API_TOKEN in config"
+    exit 0
+fi
+
+# Quick API connectivity test (5s timeout)
+HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" -m 5 \
+    -H "Authorization: Token $NETBOX_API_TOKEN" \
+    -H "Accept: application/json" \
+    "${NETBOX_API_URL}/" 2>/dev/null || echo "000")
+
+if [[ "$HTTP_CODE" == "000" ]]; then
+    echo "$PREFIX NetBox API unreachable (timeout/connection error)"
+    exit 0
+elif [[ "$HTTP_CODE" != "200" ]]; then
+    echo "$PREFIX NetBox API returned HTTP $HTTP_CODE - check credentials"
+    exit 0
+fi
+
+# Check for VMs without site assignment (data quality)
+VMS_RESPONSE=$(curl -s -m 5 \
+    -H "Authorization: Token $NETBOX_API_TOKEN" \
+    -H "Accept: application/json" \
+    "${NETBOX_API_URL}/virtualization/virtual-machines/?site__isnull=true&limit=1" 2>/dev/null || echo '{"count":0}')
+
+VMS_NO_SITE=$(echo "$VMS_RESPONSE" | grep -o '"count":[0-9]*' | cut -d: -f2 || echo "0")
+
+if [[ "$VMS_NO_SITE" -gt 0 ]]; then
+    echo "$PREFIX $VMS_NO_SITE VMs without site assignment - run /cmdb-audit for details"
+fi
+
+# Check for devices without platform
+DEVICES_RESPONSE=$(curl -s -m 5 \
+    -H "Authorization: Token $NETBOX_API_TOKEN" \
+    -H "Accept: application/json" \
+    "${NETBOX_API_URL}/dcim/devices/?platform__isnull=true&limit=1" 2>/dev/null || echo '{"count":0}')
+
+DEVICES_NO_PLATFORM=$(echo "$DEVICES_RESPONSE" | grep -o '"count":[0-9]*' | cut -d: -f2 || echo "0")
+
+if [[ "$DEVICES_NO_PLATFORM" -gt 0 ]]; then
+    echo "$PREFIX $DEVICES_NO_PLATFORM devices without platform - consider updating"
+fi
+
+exit 0

plugins/cmdb-assistant/hooks/validate-input.sh

@@ -0,0 +1,79 @@
+#!/bin/bash
+# cmdb-assistant PreToolUse validation hook
+# Validates input parameters for create/update operations
+# NON-BLOCKING: Warns but allows operation to proceed (always exits 0)
+
+set -euo pipefail
+
+PREFIX="[cmdb-assistant]"
+
+# Read tool input from stdin
+INPUT=$(cat)
+
+# Extract tool name from the input
+# Format varies, try to find tool_name or name field
+TOOL_NAME=""
+if echo "$INPUT" | grep -q '"tool_name"'; then
+    TOOL_NAME=$(echo "$INPUT" | grep -o '"tool_name"[[:space:]]*:[[:space:]]*"[^"]*"' | head -1 | sed 's/.*"\([^"]*\)"$/\1/' || true)
+elif echo "$INPUT" | grep -q '"name"'; then
+    TOOL_NAME=$(echo "$INPUT" | grep -o '"name"[[:space:]]*:[[:space:]]*"[^"]*"' | head -1 | sed 's/.*"\([^"]*\)"$/\1/' || true)
+fi
+
+# If we can't determine the tool, exit silently
+if [[ -z "$TOOL_NAME" ]]; then
+    exit 0
+fi
+
+# VM creation/update validation
+if echo "$TOOL_NAME" | grep -qE "virt_create_vm|virt_create_virtual_machine|virt_update_vm|virt_update_virtual_machine"; then
+    WARNINGS=()
+
+    # Check for missing site
+    if ! echo "$INPUT" | grep -qE '"site"[[:space:]]*:[[:space:]]*[0-9]'; then
+        WARNINGS+=("no site assigned")
+    fi
+
+    # Check for missing tenant
+    if ! echo "$INPUT" | grep -qE '"tenant"[[:space:]]*:[[:space:]]*[0-9]'; then
+        WARNINGS+=("no tenant assigned")
+    fi
+
+    # Check for missing platform
+    if ! echo "$INPUT" | grep -qE '"platform"[[:space:]]*:[[:space:]]*[0-9]'; then
+        WARNINGS+=("no platform assigned")
+    fi
+
+    if [[ ${#WARNINGS[@]} -gt 0 ]]; then
+        echo "$PREFIX VM best practice: $(IFS=', '; echo "${WARNINGS[*]}") - consider assigning for data quality"
+    fi
+fi
+
+# Device creation/update validation
+if echo "$TOOL_NAME" | grep -qE "dcim_create_device|dcim_update_device"; then
+    WARNINGS=()
+
+    # Check for missing platform
+    if ! echo "$INPUT" | grep -qE '"platform"[[:space:]]*:[[:space:]]*[0-9]'; then
+        WARNINGS+=("no platform assigned")
+    fi
+
+    # Check for missing tenant
+    if ! echo "$INPUT" | grep -qE '"tenant"[[:space:]]*:[[:space:]]*[0-9]'; then
+        WARNINGS+=("no tenant assigned")
+    fi
+
+    if [[ ${#WARNINGS[@]} -gt 0 ]]; then
+        echo "$PREFIX Device best practice: $(IFS=', '; echo "${WARNINGS[*]}") - consider assigning"
+    fi
+fi
+
+# Cluster creation validation
+if echo "$TOOL_NAME" | grep -qE "virt_create_cluster"; then
+    # Check for missing site scope
+    if ! echo "$INPUT" | grep -qE '"site"[[:space:]]*:[[:space:]]*[0-9]'; then
+        echo "$PREFIX Cluster best practice: no site scope - clusters should be scoped to a site"
+    fi
+fi
+
+# Always allow operation (non-blocking)
+exit 0

plugins/cmdb-assistant/skills/netbox-patterns/SKILL.md

@@ -0,0 +1,249 @@
+---
+description: NetBox best practices for data quality and consistency based on official NetBox Labs guidelines
+---
+
+# NetBox Best Practices Skill
+
+Reference documentation for proper NetBox data modeling, following official NetBox Labs guidelines.
+
+## CRITICAL: Dependency Order
+
+Objects must be created in this order due to foreign key dependencies. Creating objects out of order results in validation errors.
+
+```
+1. ORGANIZATION (no dependencies)
+   ├── Tenant Groups
+   ├── Tenants (optional: Tenant Group)
+   ├── Regions
+   ├── Site Groups
+   └── Tags
+
+2. SITES AND LOCATIONS
+   ├── Sites (optional: Region, Site Group, Tenant)
+   └── Locations (requires: Site, optional: parent Location)
+
+3. DCIM PREREQUISITES
+   ├── Manufacturers
+   ├── Device Types (requires: Manufacturer)
+   ├── Platforms
+   ├── Device Roles
+   └── Rack Roles
+
+4. RACKS
+   └── Racks (requires: Site, optional: Location, Rack Role, Tenant)
+
+5. DEVICES
+   ├── Devices (requires: Device Type, Role, Site; optional: Rack, Location)
+   └── Interfaces (requires: Device)
+
+6. VIRTUALIZATION
+   ├── Cluster Types
+   ├── Cluster Groups
+   ├── Clusters (requires: Cluster Type, optional: Site)
+   ├── Virtual Machines (requires: Cluster OR Site)
+   └── VM Interfaces (requires: Virtual Machine)
+
+7. IPAM
+   ├── VRFs (optional: Tenant)
+   ├── Prefixes (optional: VRF, Site, Tenant)
+   ├── IP Addresses (optional: VRF, Tenant, Interface)
+   └── VLANs (optional: Site, Tenant)
+
+8. CONNECTIONS (last)
+   └── Cables (requires: endpoints)
+```
+
+**Key Rule:** NEVER create a VM before its cluster exists. NEVER create a device before its site exists.
+
+## HIGH: Site Assignment
+
+**All infrastructure objects should have a site:**
+
+| Object Type | Site Requirement |
+|-------------|------------------|
+| Devices | **REQUIRED** |
+| Racks | **REQUIRED** |
+| VMs | RECOMMENDED (via cluster or direct) |
+| Clusters | RECOMMENDED |
+| Prefixes | RECOMMENDED |
+| VLANs | RECOMMENDED |
+
+**Why Sites Matter:**
+- Location-based queries and filtering
+- Power and capacity budgeting
+- Physical inventory tracking
+- Compliance and audit requirements
+
+## HIGH: Tenant Usage
+
+Use tenants for logical resource separation:
+
+**When to Use Tenants:**
+- Multi-team environments (assign resources to teams)
+- Multi-customer scenarios (MSP, hosting)
+- Cost allocation requirements
+- Access control boundaries
+
+**Apply Tenants To:**
+- Sites (who owns the physical location)
+- Devices (who operates the hardware)
+- VMs (who owns the workload)
+- Prefixes (who owns the IP space)
+- VLANs (who owns the network segment)
+
+## HIGH: Platform Tracking
+
+Platforms track OS/runtime information for automation and lifecycle management.
+
+**Platform Examples:**
+| Device Type | Platform Examples |
+|-------------|-------------------|
+| Servers | Ubuntu 24.04, Windows Server 2022, RHEL 9 |
+| Network | Cisco IOS 17.x, Junos 23.x, Arista EOS |
+| Raspberry Pi | Raspberry Pi OS (Bookworm), Ubuntu Server ARM |
+| Containers | Docker Container (as runtime indicator) |
+
+**Benefits:**
+- Vulnerability tracking (CVE correlation)
+- Configuration management integration
+- Lifecycle management (EOL tracking)
+- Automation targeting
+
+## MEDIUM: Tag Conventions
+
+Use tags for cross-cutting classification that spans object types.
+
+**Recommended Tag Patterns:**
+
+| Pattern | Purpose | Examples |
+|---------|---------|----------|
+| `env:*` | Environment classification | `env:production`, `env:staging`, `env:development` |
+| `app:*` | Application grouping | `app:web`, `app:database`, `app:monitoring` |
+| `team:*` | Ownership | `team:platform`, `team:infra`, `team:devops` |
+| `backup:*` | Backup policy | `backup:daily`, `backup:weekly`, `backup:none` |
+| `monitoring:*` | Monitoring level | `monitoring:critical`, `monitoring:standard` |
+
+**Tags vs Custom Fields:**
+- Tags: Cross-object classification, multiple values, filtering
+- Custom Fields: Object-specific structured data, single values, reporting
+
+## MEDIUM: Naming Conventions
+
+Consistent naming improves searchability and automation compatibility.
+
+**Recommended Patterns:**
+
+| Object Type | Pattern | Examples |
+|-------------|---------|----------|
+| Devices | `{role}-{location}-{number}` | `web-dc1-01`, `db-cloud-02`, `fw-home-01` |
+| VMs | `{env}-{app}-{number}` | `prod-api-01`, `dev-worker-03` |
+| Clusters | `{site}-{type}` | `dc1-vmware`, `home-docker` |
+| Prefixes | Include purpose in description | "Production web tier /24" |
+| VLANs | `{site}-{function}` | `dc1-mgmt`, `home-iot` |
+
+**Avoid:**
+- Inconsistent casing (mixing `HotServ` and `hotserv`)
+- Mixed separators (mixing `hhl_cluster` and `media-cluster`)
+- Generic names without context (`server1`, `vm2`)
+- Special characters other than hyphen
+
+## MEDIUM: Role Consolidation
+
+Avoid role fragmentation - use general roles with platform/tags for specificity.
+
+**Instead of:**
+```
+nginx-web-server
+apache-web-server
+web-server-frontend
+web-server-api
+```
+
+**Use:**
+```
+web-server (role) + platform (nginx/apache) + tags (frontend, api)
+```
+
+**Recommended Role Categories:**
+
+| Category | Roles |
+|----------|-------|
+| Infrastructure | `hypervisor`, `storage-server`, `network-device`, `firewall` |
+| Compute | `application-server`, `database-server`, `web-server`, `api-server` |
+| Services | `container-host`, `load-balancer`, `monitoring-server`, `backup-server` |
+| Development | `development-workstation`, `ci-runner`, `build-server` |
+| Containers | `reverse-proxy`, `database`, `cache`, `queue`, `worker` |
+
+## Docker Containers as VMs
+
+NetBox's Virtualization module can model Docker containers:
+
+**Approach:**
+1. Create device for physical Docker host
+2. Create cluster (type: "Docker Compose" or "Docker Swarm")
+3. Associate cluster with host device
+4. Create VMs for each container in the cluster
+
+**VM Fields for Containers:**
+- `name`: Container name (e.g., `media_jellyfin`)
+- `role`: Container function (e.g., `Media Server`)
+- `vcpus`: CPU limit/shares
+- `memory`: Memory limit (MB)
+- `disk`: Volume size estimate
+- `description`: Container purpose
+- `comments`: Image, ports, volumes, dependencies
+
+**This is a pragmatic modeling choice** - containers aren't VMs, but the Virtualization module is the closest fit for tracking container workloads.
+
+## Primary IP Workflow
+
+To set a device/VM's primary IP:
+
+1. Create interface on device/VM
+2. Create IP address assigned to interface
+3. Set IP as `primary_ip4` or `primary_ip6` on device/VM
+
+**Why Primary IP Matters:**
+- Used for device connectivity checks
+- Displayed in device list views
+- Used by automation tools (NAPALM, Ansible)
+- Required for many integrations
+
+## Data Quality Checklist
+
+Before closing a sprint or audit:
+
+- [ ] All VMs have site assignment (direct or via cluster)
+- [ ] All VMs have tenant assignment
+- [ ] All active devices have platform
+- [ ] All active devices have primary IP
+- [ ] Naming follows conventions
+- [ ] No orphaned prefixes (allocated but unused)
+- [ ] Tags applied consistently
+- [ ] Clusters scoped to sites
+- [ ] Roles not overly fragmented
+
+## MCP Tool Reference
+
+**Dependency Order for Creation:**
+```
+1. dcim_create_site
+2. dcim_create_manufacturer
+3. dcim_create_device_type
+4. dcim_create_device_role
+5. dcim_create_platform
+6. dcim_create_device
+7. virt_create_cluster_type
+8. virt_create_cluster
+9. virt_create_vm
+10. dcim_create_interface / virt_create_vm_interface
+11. ipam_create_ip_address
+12. dcim_update_device (set primary_ip4)
+```
+
+**Lookup Before Create:**
+Always check if object exists before creating to avoid duplicates:
+```
+1. dcim_list_devices name=<hostname>
+2. If exists, update; if not, create
+```

plugins/contract-validator/.mcp.json

@@ -2,9 +2,8 @@
   "mcpServers": {
     "contract-validator": {
       "type": "stdio",
-      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/contract-validator/.venv/bin/python",
-      "args": ["-m", "mcp_server.server"],
-      "cwd": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/contract-validator"
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/contract-validator/run.sh",
+      "args": []
     }
   }
 }

plugins/contract-validator/README.md

@@ -40,6 +40,7 @@ pip install -r requirements.txt
 
 | Command | Description |
 |---------|-------------|
+| `/initial-setup` | Interactive setup wizard |
 | `/validate-contracts` | Full marketplace compatibility validation |
 | `/check-agent` | Validate single agent definition |
 | `/list-interfaces` | Show all plugin interfaces |

plugins/contract-validator/commands/initial-setup.md

@@ -0,0 +1,152 @@
+---
+description: Interactive setup wizard for contract-validator plugin - verifies MCP server and shows capabilities
+---
+
+# Contract-Validator Setup Wizard
+
+This command sets up the contract-validator plugin for cross-plugin compatibility validation.
+
+## Important Context
+
+- **This command uses Bash, Read, Write, and AskUserQuestion tools** - NOT MCP tools
+- **MCP tools won't work until after setup + session restart**
+- **No external credentials required** - this plugin validates local files only
+
+---
+
+## Phase 1: Environment Validation
+
+### Step 1.1: Check Python Version
+
+```bash
+python3 --version
+```
+
+Requires Python 3.10+. If below, stop setup and inform user:
+```
+Python 3.10 or higher is required. Please install it and run /initial-setup again.
+```
+
+---
+
+## Phase 2: MCP Server Setup
+
+### Step 2.1: Locate Contract-Validator MCP Server
+
+```bash
+# If running from installed marketplace
+ls -la ~/.claude/plugins/marketplaces/leo-claude-mktplace/mcp-servers/contract-validator/ 2>/dev/null || echo "NOT_FOUND_INSTALLED"
+
+# If running from source
+ls -la ~/claude-plugins-work/mcp-servers/contract-validator/ 2>/dev/null || echo "NOT_FOUND_SOURCE"
+```
+
+Determine which path exists and use that as the MCP server path.
+
+### Step 2.2: Check Virtual Environment
+
+```bash
+ls -la /path/to/mcp-servers/contract-validator/.venv/bin/python 2>/dev/null && echo "VENV_EXISTS" || echo "VENV_MISSING"
+```
+
+### Step 2.3: Create Virtual Environment (if missing)
+
+```bash
+cd /path/to/mcp-servers/contract-validator && python3 -m venv .venv && source .venv/bin/activate && pip install --upgrade pip && pip install -r requirements.txt && deactivate
+```
+
+**If pip install fails:**
+- Show the error to the user
+- Suggest: "Check your internet connection and try again."
+
+---
+
+## Phase 3: Validation
+
+### Step 3.1: Verify MCP Server
+
+```bash
+cd /path/to/mcp-servers/contract-validator && .venv/bin/python -c "from mcp_server.server import ContractValidatorMCPServer; print('MCP Server OK')"
+```
+
+If this fails, check the error and report it to the user.
+
+### Step 3.2: Summary
+
+Display:
+
+```
+╔════════════════════════════════════════════════════════════════╗
+║          CONTRACT-VALIDATOR SETUP COMPLETE                      ║
+╠════════════════════════════════════════════════════════════════╣
+║ MCP Server:        ✓ Ready                                      ║
+║ Parse Tools:       ✓ Available (2 tools)                        ║
+║ Validation Tools:  ✓ Available (3 tools)                        ║
+║ Report Tools:      ✓ Available (2 tools)                        ║
+╚════════════════════════════════════════════════════════════════╝
+```
+
+### Step 3.3: Session Restart Notice
+
+---
+
+**Session Restart Required**
+
+Restart your Claude Code session for MCP tools to become available.
+
+**After restart, you can:**
+- Run `/validate-contracts` to check all plugins for compatibility issues
+- Run `/check-agent` to validate a single agent definition
+- Run `/list-interfaces` to see all plugin commands and tools
+
+---
+
+## Available Tools
+
+| Category | Tools | Description |
+|----------|-------|-------------|
+| Parse | `parse_plugin_interface`, `parse_claude_md_agents` | Extract interfaces from README.md and agents from CLAUDE.md |
+| Validation | `validate_compatibility`, `validate_agent_refs`, `validate_data_flow` | Check conflicts, tool references, and data flows |
+| Report | `generate_compatibility_report`, `list_issues` | Generate reports and filter issues |
+
+---
+
+## Available Commands
+
+| Command | Description |
+|---------|-------------|
+| `/validate-contracts` | Full marketplace compatibility validation |
+| `/check-agent` | Validate single agent definition |
+| `/list-interfaces` | Show all plugin interfaces |
+
+---
+
+## Use Cases
+
+### 1. Pre-Release Validation
+Run `/validate-contracts` before releasing a new marketplace version to catch:
+- Command name conflicts between plugins
+- Missing tool references in agents
+- Broken data flows
+
+### 2. Agent Development
+Run `/check-agent` when creating or modifying agents to verify:
+- All referenced tools exist
+- Data flows are valid
+- No undeclared dependencies
+
+### 3. Plugin Audit
+Run `/list-interfaces` to get a complete view of:
+- All commands across plugins
+- All tools available
+- Potential overlap areas
+
+---
+
+## No Configuration Required
+
+This plugin doesn't require any configuration files. It reads plugin manifests and README files directly from the filesystem.
+
+**Paths it scans:**
+- Marketplace: `~/.claude/plugins/marketplaces/leo-claude-mktplace/plugins/`
+- Source (if available): `~/claude-plugins-work/plugins/`

plugins/data-platform/.mcp.json

@@ -2,9 +2,8 @@
   "mcpServers": {
     "data-platform": {
       "type": "stdio",
-      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/data-platform/.venv/bin/python",
-      "args": ["-m", "mcp_server.server"],
-      "cwd": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/data-platform"
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/data-platform/run.sh",
+      "args": []
     }
   }
 }

plugins/pr-review/.mcp.json

@@ -1,12 +1,8 @@
 {
   "mcpServers": {
     "gitea": {
-      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea/.venv/bin/python",
-      "args": ["-m", "mcp_server.server"],
-      "cwd": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea",
-      "env": {
-        "PYTHONPATH": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea"
-      }
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea/run.sh",
+      "args": []
     }
   }
 }

plugins/projman/.mcp.json

@@ -1,12 +1,8 @@
 {
   "mcpServers": {
     "gitea": {
-      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea/.venv/bin/python",
-      "args": ["-m", "mcp_server.server"],
-      "cwd": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea",
-      "env": {
-        "PYTHONPATH": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea"
-      }
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/gitea/run.sh",
+      "args": []
     }
   }
 }

plugins/projman/hooks/startup-check.sh

@@ -5,13 +5,43 @@
 
 PREFIX="[projman]"
 
-# Check if MCP venv exists
+# Calculate paths
 PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-$(dirname "$(dirname "$(realpath "$0")")")}"
-VENV_PATH="$PLUGIN_ROOT/mcp-servers/gitea/.venv/bin/python"
+# Marketplace root is 2 levels up from plugin root (plugins/projman -> .)
+MARKETPLACE_ROOT="$(dirname "$(dirname "$PLUGIN_ROOT")")"
+VENV_REPAIR_SCRIPT="$MARKETPLACE_ROOT/scripts/venv-repair.sh"
+PLUGIN_CACHE="$HOME/.claude/plugins/cache/leo-claude-mktplace"
 
-if [[ ! -f "$VENV_PATH" ]]; then
-    echo "$PREFIX MCP venvs missing - run setup.sh from installed marketplace"
-    exit 0
+# ============================================================================
+# Clear stale plugin cache (MUST run before MCP servers load)
+# ============================================================================
+# The cache at ~/.claude/plugins/cache/ holds versioned .mcp.json files.
+# After marketplace updates, cached configs may point to old paths.
+# Clearing forces Claude to read fresh configs from installed marketplace.
+
+if [[ -d "$PLUGIN_CACHE" ]]; then
+    rm -rf "$PLUGIN_CACHE"
+    # Don't output anything - this should be silent and automatic
+fi
+
+# ============================================================================
+# Auto-repair MCP venvs (runs before other checks)
+# ============================================================================
+
+if [[ -x "$VENV_REPAIR_SCRIPT" ]]; then
+    # Run venv repair - this creates symlinks to cached venvs
+    # Only outputs messages if something needed fixing
+    "$VENV_REPAIR_SCRIPT" 2>/dev/null || {
+        echo "$PREFIX MCP venv setup failed - run: cd $MARKETPLACE_ROOT && ./scripts/setup-venvs.sh"
+        exit 0
+    }
+else
+    # Fallback: just check if venv exists
+    VENV_PATH="$PLUGIN_ROOT/mcp-servers/gitea/.venv/bin/python"
+    if [[ ! -f "$VENV_PATH" ]]; then
+        echo "$PREFIX MCP venvs missing - run setup.sh from installed marketplace"
+        exit 0
+    fi
 fi
 
 # Check git remote vs .env config (only if .env exists)

plugins/viz-platform/.mcp.json

@@ -2,9 +2,8 @@
   "mcpServers": {
     "viz-platform": {
       "type": "stdio",
-      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/viz-platform/.venv/bin/python",
-      "args": ["-m", "mcp_server.server"],
-      "cwd": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/viz-platform"
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-servers/viz-platform/run.sh",
+      "args": []
     }
   }
 }

scripts/post-update.sh

@@ -1,17 +1,20 @@
 #!/usr/bin/env bash
 #
-# post-update.sh - Run after pulling updates
+# post-update.sh - Run after pulling updates or marketplace sync
 #
 # Usage: ./scripts/post-update.sh
 #
 # This script:
-# 1. Updates Python dependencies for MCP servers
-# 2. Validates configuration still works
-# 3. Reports any new manual steps from CHANGELOG
+# 1. Clears Claude plugin cache (forces fresh .mcp.json reads)
+# 2. Restores MCP venv symlinks (instant if cache exists)
+# 3. Creates venvs in external cache if missing (first run only)
+# 4. Shows recent changelog updates
 #
 
 set -euo pipefail
 
+CLAUDE_PLUGIN_CACHE="$HOME/.claude/plugins/cache/leo-claude-mktplace"
+
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 REPO_ROOT="$(dirname "$SCRIPT_DIR")"
 
@@ -19,42 +22,25 @@ REPO_ROOT="$(dirname "$SCRIPT_DIR")"
 GREEN='\033[0;32m'
 YELLOW='\033[1;33m'
 BLUE='\033[0;34m'
+RED='\033[0;31m'
 NC='\033[0m'
 
 log_info() { echo -e "${BLUE}[INFO]${NC} $1"; }
 log_success() { echo -e "${GREEN}[OK]${NC} $1"; }
 log_warn() { echo -e "${YELLOW}[WARN]${NC} $1"; }
-
-update_mcp_server() {
-    local server_name="$1"
-    local server_path="$REPO_ROOT/mcp-servers/$server_name"
-
-    log_info "Updating $server_name dependencies..."
-
-    if [[ -d "$server_path/.venv" ]] && [[ -f "$server_path/requirements.txt" ]]; then
-        cd "$server_path"
-        source .venv/bin/activate
-        pip install -q --upgrade pip
-        pip install -q -r requirements.txt
-        deactivate
-        cd "$REPO_ROOT"
-        log_success "$server_name dependencies updated"
-    else
-        log_warn "$server_name not fully set up - run ./scripts/setup.sh first"
-    fi
-}
+log_error() { echo -e "${RED}[ERROR]${NC} $1"; }
 
 check_changelog() {
-    log_info "Checking CHANGELOG for recent updates..."
-
     if [[ -f "$REPO_ROOT/CHANGELOG.md" ]]; then
-        # Show the Unreleased section
-        echo ""
-        echo "Recent changes (from CHANGELOG.md):"
-        echo "-----------------------------------"
-        sed -n '/## \[Unreleased\]/,/## \[/p' "$REPO_ROOT/CHANGELOG.md" | head -30
-        echo "-----------------------------------"
-        echo ""
+        local unreleased
+        unreleased=$(sed -n '/## \[Unreleased\]/,/## \[/p' "$REPO_ROOT/CHANGELOG.md" | grep -E '^### ' | head -1 || true)
+        if [[ -n "$unreleased" ]]; then
+            echo ""
+            log_info "Recent changes (from CHANGELOG.md):"
+            echo "-----------------------------------"
+            sed -n '/## \[Unreleased\]/,/## \[/p' "$REPO_ROOT/CHANGELOG.md" | head -20
+            echo "-----------------------------------"
+        fi
     fi
 }
 
@@ -64,23 +50,37 @@ main() {
     echo "=============================================="
     echo ""
 
-    # Shared MCP servers at repository root (v3.0.0+)
-    update_mcp_server "gitea"
-    update_mcp_server "netbox"
-    update_mcp_server "data-platform"
+    # Clear Claude plugin cache to force fresh .mcp.json reads
+    # This cache holds versioned copies that become stale after updates
+    if [[ -d "$CLAUDE_PLUGIN_CACHE" ]]; then
+        log_info "Clearing Claude plugin cache..."
+        rm -rf "$CLAUDE_PLUGIN_CACHE"
+        log_success "Plugin cache cleared"
+    fi
+
+    # Run venv-repair.sh to restore symlinks to external cache
+    # This is instant if cache exists, or does full setup on first run
+    if [[ -x "$SCRIPT_DIR/venv-repair.sh" ]]; then
+        log_info "Restoring MCP venv symlinks..."
+        if "$SCRIPT_DIR/venv-repair.sh"; then
+            log_success "MCP venvs ready"
+        else
+            log_error "MCP venv setup failed"
+            log_warn "Run: $SCRIPT_DIR/setup-venvs.sh for full setup"
+            exit 1
+        fi
+    else
+        log_error "venv-repair.sh not found at $SCRIPT_DIR"
+        exit 1
+    fi
 
     check_changelog
 
     echo ""
     log_success "Post-update complete!"
     echo ""
-    echo "If you see new features in the changelog that require"
-    echo "configuration changes, update your ~/.config/claude/*.env files."
+    echo "IMPORTANT: Restart Claude Code for changes to take effect."
+    echo "MCP servers will work immediately on next session start."
 }
 
 main "$@"
-
-# Clear plugin cache to ensure fresh hooks are loaded
-echo "Clearing plugin cache..."
-rm -rf ~/.claude/plugins/cache/leo-claude-mktplace/
-echo "Cache cleared"

scripts/setup-venvs.sh

@@ -0,0 +1,281 @@
+#!/usr/bin/env bash
+#
+# setup-venvs.sh - Smart MCP server venv management with external cache
+#
+# This script manages Python virtual environments for MCP servers in a
+# PERSISTENT location outside the marketplace directory, so they survive
+# marketplace updates.
+#
+# Features:
+# - Stores venvs in ~/.cache/claude-mcp-venvs/ (survives updates)
+# - Incremental installs (only missing packages)
+# - Hash-based change detection (skip if requirements unchanged)
+# - Can be called from SessionStart hooks safely
+#
+# Usage:
+#   ./scripts/setup-venvs.sh              # Full setup
+#   ./scripts/setup-venvs.sh --check      # Check only, no install
+#   ./scripts/setup-venvs.sh --quick      # Skip if hash unchanged
+#   ./scripts/setup-venvs.sh gitea        # Setup specific server only
+#
+
+set -euo pipefail
+
+# ============================================================================
+# Configuration
+# ============================================================================
+
+# Persistent venv location (outside marketplace)
+VENV_CACHE_DIR="${HOME}/.cache/claude-mcp-venvs/leo-claude-mktplace"
+
+# Script and repo paths
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+REPO_ROOT="$(dirname "$SCRIPT_DIR")"
+
+# MCP servers to manage
+MCP_SERVERS=(gitea netbox data-platform viz-platform contract-validator)
+
+# Colors
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+BLUE='\033[0;34m'
+NC='\033[0m'
+
+# Flags
+CHECK_ONLY=false
+QUICK_MODE=false
+SPECIFIC_SERVER=""
+
+# ============================================================================
+# Argument Parsing
+# ============================================================================
+
+while [[ $# -gt 0 ]]; do
+    case $1 in
+        --check)
+            CHECK_ONLY=true
+            shift
+            ;;
+        --quick)
+            QUICK_MODE=true
+            shift
+            ;;
+        -h|--help)
+            echo "Usage: $0 [OPTIONS] [SERVER]"
+            echo ""
+            echo "Options:"
+            echo "  --check    Check venv status without installing"
+            echo "  --quick    Skip servers with unchanged requirements"
+            echo "  -h,--help  Show this help"
+            echo ""
+            echo "Servers: ${MCP_SERVERS[*]}"
+            exit 0
+            ;;
+        *)
+            SPECIFIC_SERVER="$1"
+            shift
+            ;;
+    esac
+done
+
+# ============================================================================
+# Helper Functions
+# ============================================================================
+
+log_info() { echo -e "${BLUE}[INFO]${NC} $1"; }
+log_ok() { echo -e "${GREEN}[OK]${NC} $1"; }
+log_skip() { echo -e "${YELLOW}[SKIP]${NC} $1"; }
+log_error() { echo -e "${RED}[ERROR]${NC} $1" >&2; }
+log_warn() { echo -e "${YELLOW}[WARN]${NC} $1"; }
+
+# Calculate hash of requirements file(s)
+requirements_hash() {
+    local server_path="$1"
+    local hash_input=""
+
+    if [[ -f "$server_path/requirements.txt" ]]; then
+        hash_input+=$(cat "$server_path/requirements.txt")
+    fi
+    if [[ -f "$server_path/pyproject.toml" ]]; then
+        hash_input+=$(cat "$server_path/pyproject.toml")
+    fi
+
+    echo "$hash_input" | sha256sum | cut -d' ' -f1
+}
+
+# Check if requirements changed since last install
+requirements_changed() {
+    local server_name="$1"
+    local server_path="$2"
+    local hash_file="$VENV_CACHE_DIR/$server_name/.requirements_hash"
+
+    local current_hash
+    current_hash=$(requirements_hash "$server_path")
+
+    if [[ -f "$hash_file" ]]; then
+        local stored_hash
+        stored_hash=$(cat "$hash_file")
+        if [[ "$current_hash" == "$stored_hash" ]]; then
+            return 1  # Not changed
+        fi
+    fi
+    return 0  # Changed or no hash file
+}
+
+# Save requirements hash after successful install
+save_requirements_hash() {
+    local server_name="$1"
+    local server_path="$2"
+    local hash_file="$VENV_CACHE_DIR/$server_name/.requirements_hash"
+
+    requirements_hash "$server_path" > "$hash_file"
+}
+
+# ============================================================================
+# Main Setup Function
+# ============================================================================
+
+setup_server() {
+    local server_name="$1"
+    local server_path="$REPO_ROOT/mcp-servers/$server_name"
+    local venv_path="$VENV_CACHE_DIR/$server_name/.venv"
+
+    # Verify server exists in repo
+    if [[ ! -d "$server_path" ]]; then
+        log_error "$server_name: source directory not found at $server_path"
+        return 1
+    fi
+
+    # Check-only mode
+    if [[ "$CHECK_ONLY" == true ]]; then
+        if [[ -f "$venv_path/bin/python" ]]; then
+            log_ok "$server_name: venv exists"
+        else
+            log_error "$server_name: venv MISSING"
+            return 1
+        fi
+        return 0
+    fi
+
+    # Quick mode: skip if requirements unchanged
+    if [[ "$QUICK_MODE" == true ]] && [[ -f "$venv_path/bin/python" ]]; then
+        if ! requirements_changed "$server_name" "$server_path"; then
+            log_skip "$server_name: requirements unchanged"
+            return 0
+        fi
+    fi
+
+    log_info "$server_name: setting up venv..."
+
+    # Create cache directory
+    mkdir -p "$VENV_CACHE_DIR/$server_name"
+
+    # Create venv if missing
+    if [[ ! -d "$venv_path" ]]; then
+        python3 -m venv "$venv_path"
+        log_ok "$server_name: venv created"
+    fi
+
+    # Activate and install
+    # shellcheck disable=SC1091
+    source "$venv_path/bin/activate"
+
+    # Upgrade pip quietly
+    pip install -q --upgrade pip
+
+    # Install requirements (incremental - pip handles already-installed)
+    if [[ -f "$server_path/requirements.txt" ]]; then
+        pip install -q -r "$server_path/requirements.txt"
+    fi
+
+    # Install local package in editable mode if pyproject.toml exists
+    if [[ -f "$server_path/pyproject.toml" ]]; then
+        pip install -q -e "$server_path"
+        log_ok "$server_name: package installed (editable)"
+    fi
+
+    deactivate
+
+    # Save hash for quick mode
+    save_requirements_hash "$server_name" "$server_path"
+
+    log_ok "$server_name: ready"
+}
+
+# ============================================================================
+# Create Symlinks (for backward compatibility)
+# ============================================================================
+
+create_symlinks() {
+    log_info "Creating symlinks for backward compatibility..."
+
+    for server_name in "${MCP_SERVERS[@]}"; do
+        local server_path="$REPO_ROOT/mcp-servers/$server_name"
+        local venv_path="$VENV_CACHE_DIR/$server_name/.venv"
+        local link_path="$server_path/.venv"
+
+        # Skip if source doesn't exist
+        [[ ! -d "$server_path" ]] && continue
+
+        # Skip if venv not in cache
+        [[ ! -d "$venv_path" ]] && continue
+
+        # Remove existing venv or symlink
+        if [[ -L "$link_path" ]]; then
+            rm "$link_path"
+        elif [[ -d "$link_path" ]]; then
+            log_warn "$server_name: removing old venv directory (now using cache)"
+            rm -rf "$link_path"
+        fi
+
+        # Create symlink
+        ln -s "$venv_path" "$link_path"
+        log_ok "$server_name: symlink created"
+    done
+}
+
+# ============================================================================
+# Main
+# ============================================================================
+
+main() {
+    echo "=============================================="
+    echo "  MCP Server Venv Manager"
+    echo "=============================================="
+    echo "Cache: $VENV_CACHE_DIR"
+    echo ""
+
+    local failed=0
+
+    if [[ -n "$SPECIFIC_SERVER" ]]; then
+        # Setup specific server
+        if setup_server "$SPECIFIC_SERVER"; then
+            : # success
+        else
+            failed=1
+        fi
+    else
+        # Setup all servers
+        for server in "${MCP_SERVERS[@]}"; do
+            if ! setup_server "$server"; then
+                ((failed++)) || true
+            fi
+        done
+    fi
+
+    # Create symlinks for backward compatibility
+    if [[ "$CHECK_ONLY" != true ]]; then
+        create_symlinks
+    fi
+
+    echo ""
+    if [[ $failed -eq 0 ]]; then
+        log_ok "All MCP servers ready"
+    else
+        log_error "$failed server(s) failed"
+        return 1
+    fi
+}
+
+main "$@"

scripts/setup.sh

@@ -72,11 +72,16 @@ setup_shared_mcp() {
         log_success "$server_name venv created"
     fi
 
-    # Install/update dependencies
+    # Install/update dependencies and local package
     if [[ -f "requirements.txt" ]]; then
         source .venv/bin/activate
         pip install -q --upgrade pip
         pip install -q -r requirements.txt
+        # Install local package in editable mode (required for MCP server to work)
+        if [[ -f "pyproject.toml" ]]; then
+            pip install -q -e .
+            log_success "$server_name package installed (editable mode)"
+        fi
         deactivate
         log_success "$server_name dependencies installed"
     else
@@ -125,6 +130,24 @@ verify_symlinks() {
         log_error "data-platform symlink missing"
         log_todo "Run: ln -s ../../../mcp-servers/data-platform plugins/data-platform/mcp-servers/data-platform"
     fi
+
+    # Check viz-platform -> viz-platform symlink
+    local vizplatform_link="$REPO_ROOT/plugins/viz-platform/mcp-servers/viz-platform"
+    if [[ -L "$vizplatform_link" ]]; then
+        log_success "viz-platform symlink exists"
+    else
+        log_error "viz-platform symlink missing"
+        log_todo "Run: ln -s ../../../mcp-servers/viz-platform plugins/viz-platform/mcp-servers/viz-platform"
+    fi
+
+    # Check contract-validator -> contract-validator symlink
+    local contractvalidator_link="$REPO_ROOT/plugins/contract-validator/mcp-servers/contract-validator"
+    if [[ -L "$contractvalidator_link" ]]; then
+        log_success "contract-validator symlink exists"
+    else
+        log_error "contract-validator symlink missing"
+        log_todo "Run: ln -s ../../../mcp-servers/contract-validator plugins/contract-validator/mcp-servers/contract-validator"
+    fi
 }
 
 # --- Section 3: Config File Templates ---
@@ -301,7 +324,7 @@ print_report() {
 # --- Main ---
 main() {
     echo "=============================================="
-    echo "  Leo Claude Marketplace Setup (v3.0.0)"
+    echo "  Leo Claude Marketplace Setup (v5.0.0)"
     echo "=============================================="
     echo ""
 
@@ -309,6 +332,8 @@ main() {
     setup_shared_mcp "gitea"
     setup_shared_mcp "netbox"
     setup_shared_mcp "data-platform"
+    setup_shared_mcp "viz-platform"
+    setup_shared_mcp "contract-validator"
 
     # Verify symlinks from plugins to shared MCP servers
     verify_symlinks

scripts/venv-repair.sh

@@ -0,0 +1,169 @@
+#!/usr/bin/env bash
+#
+# venv-repair.sh - Fast MCP venv auto-repair for SessionStart hooks
+#
+# This script is designed to run at session start. It:
+# 1. Checks if venvs exist in external cache (~/.cache/claude-mcp-venvs/)
+# 2. Creates symlinks from marketplace to cache (instant operation)
+# 3. Only runs pip install if cache is missing (first install)
+#
+# Output format: All messages prefixed with [mcp-venv] for hook display
+#
+# Usage:
+#   ./scripts/venv-repair.sh           # Auto-repair (default)
+#   ./scripts/venv-repair.sh --silent  # Silent mode (no output unless error)
+#
+
+set -euo pipefail
+
+# ============================================================================
+# Configuration
+# ============================================================================
+
+PREFIX="[mcp-venv]"
+VENV_CACHE_DIR="${HOME}/.cache/claude-mcp-venvs/leo-claude-mktplace"
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+REPO_ROOT="$(dirname "$SCRIPT_DIR")"
+
+# MCP servers
+MCP_SERVERS=(gitea netbox data-platform viz-platform contract-validator)
+
+# Parse args
+SILENT=false
+[[ "${1:-}" == "--silent" ]] && SILENT=true
+
+log() {
+    [[ "$SILENT" == true ]] && return
+    echo "$PREFIX $1"
+}
+
+log_error() {
+    echo "$PREFIX ERROR: $1" >&2
+}
+
+# ============================================================================
+# Check if all venvs exist in cache
+# ============================================================================
+
+cache_complete() {
+    for server in "${MCP_SERVERS[@]}"; do
+        local venv_python="$VENV_CACHE_DIR/$server/.venv/bin/python"
+        [[ ! -f "$venv_python" ]] && return 1
+    done
+    return 0
+}
+
+# ============================================================================
+# Create symlinks from marketplace to cache
+# ============================================================================
+
+create_symlink() {
+    local server_name="$1"
+    local server_path="$REPO_ROOT/mcp-servers/$server_name"
+    local venv_cache="$VENV_CACHE_DIR/$server_name/.venv"
+    local venv_link="$server_path/.venv"
+
+    # Skip if server doesn't exist
+    [[ ! -d "$server_path" ]] && return 0
+
+    # Skip if cache doesn't exist
+    [[ ! -d "$venv_cache" ]] && return 1
+
+    # Already correct symlink?
+    if [[ -L "$venv_link" ]]; then
+        local target
+        target=$(readlink "$venv_link")
+        [[ "$target" == "$venv_cache" ]] && return 0
+        rm "$venv_link"
+    elif [[ -d "$venv_link" ]]; then
+        # Old venv directory exists - back it up or remove
+        rm -rf "$venv_link"
+    fi
+
+    # Create symlink
+    ln -s "$venv_cache" "$venv_link"
+    return 0
+}
+
+create_all_symlinks() {
+    local created=0
+    for server in "${MCP_SERVERS[@]}"; do
+        if create_symlink "$server"; then
+            ((created++)) || true
+        fi
+    done
+    [[ $created -gt 0 ]] && log "Restored $created venv symlinks"
+}
+
+# ============================================================================
+# Full setup (only if cache missing)
+# ============================================================================
+
+setup_server() {
+    local server_name="$1"
+    local server_path="$REPO_ROOT/mcp-servers/$server_name"
+    local venv_path="$VENV_CACHE_DIR/$server_name/.venv"
+
+    [[ ! -d "$server_path" ]] && return 0
+
+    mkdir -p "$VENV_CACHE_DIR/$server_name"
+
+    # Create venv
+    if [[ ! -d "$venv_path" ]]; then
+        python3 -m venv "$venv_path"
+    fi
+
+    # Install dependencies
+    # shellcheck disable=SC1091
+    source "$venv_path/bin/activate"
+    pip install -q --upgrade pip
+
+    if [[ -f "$server_path/requirements.txt" ]]; then
+        pip install -q -r "$server_path/requirements.txt"
+    fi
+
+    if [[ -f "$server_path/pyproject.toml" ]]; then
+        pip install -q -e "$server_path"
+    fi
+
+    deactivate
+
+    # Save hash for future quick checks
+    local hash_file="$VENV_CACHE_DIR/$server_name/.requirements_hash"
+    {
+        if [[ -f "$server_path/requirements.txt" ]]; then
+            cat "$server_path/requirements.txt"
+        fi
+        if [[ -f "$server_path/pyproject.toml" ]]; then
+            cat "$server_path/pyproject.toml"
+        fi
+        echo ""  # Ensure non-empty input for sha256sum
+    } | sha256sum | cut -d' ' -f1 > "$hash_file"
+}
+
+full_setup() {
+    log "First run - setting up MCP venvs (this only happens once)..."
+    for server in "${MCP_SERVERS[@]}"; do
+        log "  Setting up $server..."
+        setup_server "$server"
+    done
+    log "Setup complete. Future sessions will be instant."
+}
+
+# ============================================================================
+# Main
+# ============================================================================
+
+main() {
+    # Fast path: cache exists, just ensure symlinks
+    if cache_complete; then
+        create_all_symlinks
+        exit 0
+    fi
+
+    # Slow path: need to create venvs (first install)
+    full_setup
+    create_all_symlinks
+}
+
+main "$@"

scripts/verify-hooks.sh

@@ -23,7 +23,7 @@ if [ -d ~/.claude/plugins/cache/leo-claude-mktplace ]; then
 fi
 
 # Verify installed hooks are command type
-for plugin in doc-guardian code-sentinel projman pr-review project-hygiene data-platform; do
+for plugin in doc-guardian code-sentinel projman pr-review project-hygiene data-platform cmdb-assistant; do
     HOOK_FILE=~/.claude/plugins/marketplaces/leo-claude-mktplace/plugins/$plugin/hooks/hooks.json
     if [ -f "$HOOK_FILE" ]; then
         if grep -q '"type": "command"' "$HOOK_FILE" || grep -q '"type":"command"' "$HOOK_FILE"; then