Back to skills
SKILL.md
Ansible Mcp Schema Lookup
ASecurityWorkflow for querying the Ansible MCP server to discover available tools, inspect resource schemas, access documentation endpoints, and validate definitions before code generation. Use when: (1) discovering available MCP tools and their parameter signatures via list_available_tools, (2) querying zen_of_ansible or ansible_content_best_practices for philosophy and conventions, (3) reading guidelines://ansible-content-best-practices resource URI for deep project structure standards, (4) validati...
- 2 stars
- 0 votes
- 0 copies
- 1 view
- Added September 29, 2026
Works with
Security analysis
100/100npx -y skills add talkops-ai/opscode --skill ansible-mcp-schema-lookup --agent claude-codeAre you the author of Ansible Mcp Schema Lookup?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/talkops-ai-ansible-mcp-schema-lookup)---
name: ansible-mcp-schema-lookup
description: >-
Workflow for querying the Ansible MCP server to discover available tools,
inspect resource schemas, access documentation endpoints, and validate
definitions before code generation. Use when: (1) discovering available MCP
tools and their parameter signatures via list_available_tools, (2) querying
zen_of_ansible or ansible_content_best_practices for philosophy and conventions,
(3) reading guidelines://ansible-content-best-practices resource URI for deep
project structure standards, (4) validating execution-environment.yml against
schema://execution-environment and rules://execution-environment MCP resources,
(5) inspecting ade_environment_info response fields before environment setup,
(6) checking ade_setup_environment parameter schemas (envName, installRequirements,
collections), or (7) resolving MCP tool invocation errors. Do NOT use for writing
playbook code (use ansible-code-authoring), linting (use ansible-linting-remediation),
or deployment execution (use ansible-runner-execution).
license: MIT
compatibility: designed for opscode
---
# Ansible MCP Schema Lookup
Query the Ansible Development Tools MCP server to discover tools, inspect parameter schemas, access documentation endpoints, and validate definitions before writing or modifying Ansible configurations.
---
## Core Principles
1. **Schema-First**: Never invoke an MCP tool without first understanding its parameter schema. Guessing parameters leads to tool invocation failures.
2. **Ground Before Generate**: Always query documentation endpoints (`zen_of_ansible`, `ansible_content_best_practices`) before authoring complex playbooks.
3. **Validate Against Schema**: Before building Execution Environments, validate definitions against the MCP schema and rules resources.
4. **Resource URIs Are Read-Only**: MCP resources (`guidelines://`, `schema://`, `rules://`) are read-only references — they cannot be modified.
---
## MCP Server Connection
The Ansible MCP server connects via `stdio` transport:
```json
{
"mcpServers": {
"ansible": {
"command": "npx",
"args": ["-y", "@ansible/ansible-mcp-server", "--stdio"],
"env": {
"WORKSPACE_ROOT": "."
}
}
}
}
```
`WORKSPACE_ROOT` scopes all file operations to the designated workspace directory.
---
## Available Tools
### Environment & Scaffolding
| MCP Tool | Purpose | Key Parameters |
|---|---|---|
| `ade_environment_info` | Discover host system state — Python version, Ansible Core version, virtualenv status, installed collections | None required |
| `ade_setup_environment` | Provision isolated Python venv, install core tooling and collection dependencies | `envName: "venv"`, `installRequirements: true`, `collections: ["community.general"]` |
| `ansible_create_collection` | Scaffold collection structure with galaxy.yml, plugins/, roles/ | `namespace`, `name`, `path` |
| `ansible_create_playbook` | Scaffold playbook project with inventory/, playbooks/, roles/ | `name`, `path` |
### Knowledge Base & Documentation
| MCP Tool / Resource | Type | Purpose |
|---|---|---|
| `zen_of_ansible` | Tool endpoint | Core philosophical tenets — "Playbooks are not for programming", declarative state > imperative logic |
| `ansible_content_best_practices` | Tool endpoint | Naming conventions, structural organisation, modularity standards |
| `guidelines://ansible-content-best-practices` | Resource URI (read-only) | Deep-dive project structure standards, testing strategies, naming patterns |
### Validation & Execution
| MCP Tool / Resource | Type | Purpose |
|---|---|---|
| `ansible_lint` | Tool endpoint | Static analysis with profiles (min → production), auto-fix (`fix: true`), SARIF/JSON output |
| `ansible_navigator` | Tool endpoint | Production playbook execution inside Execution Environments |
| `define_and_build_execution_env` | Tool endpoint | Build containerised EE images via Ansible Builder v3 |
| `schema://execution-environment` | Resource URI (read-only) | JSON Schema for validating `execution-environment.yml` definitions |
| `rules://execution-environment` | Resource URI (read-only) | Validation rules and constraints for EE definitions |
---
## Execution Workflow
### Step 1: Discover Available Tools
Before any operation, verify tool availability:
1. List all available MCP tools via the server
2. Inspect parameter schemas for each tool you intend to use
3. Confirm required vs optional parameters
### Step 2: Query Documentation Endpoints
Before authoring playbooks or collections:
1. **`zen_of_ansible`** — Ingest core Ansible philosophy
2. **`ansible_content_best_practices`** — Retrieve naming, structure, modularity standards
3. **`guidelines://ansible-content-best-practices`** — Deep-dive resource URI for complex refactoring
### Step 3: Inspect Environment Discovery Fields
Before scaffolding or executing:
1. Call `ade_environment_info` to inspect the workspace
2. Check response fields: Python version, Ansible Core version, virtualenv status, installed collections
3. If tooling is missing, call `ade_setup_environment` with:
- `envName: "venv"` — creates an isolated Python venv
- `installRequirements: true` — installs from workspace `requirements.yml` / `requirements.txt`
- `collections: [...]` — array of collections to install
### Step 4: Validate EE Definitions Against Schema
Before building Execution Environments:
1. Read `schema://execution-environment` to understand the Version 3 schema structure
2. Read `rules://execution-environment` to understand validation constraints
3. Validate the authored `execution-environment.yml` against both resources
4. Invoke `define_and_build_execution_env` only after validation passes
### Step 5: Inspect Lint Tool Parameters
Before running linting:
1. Confirm `ansible_lint` supports the target profile (`min`, `basic`, `moderate`, `safety`, `shared`, `production`)
2. Confirm `fix: true` parameter is available for auto-remediation
3. Check output format options (SARIF, JSON) for structured violation parsing
---
## Quick Reference Map
| Workflow Step | What to Check | MCP Query |
|---|---|---|
| **Tool Discovery** | Available tools and parameter schemas | List available tools from MCP server |
| **Philosophy Grounding** | Ansible design tenets | `zen_of_ansible` |
| **Best Practices** | Naming, structure, modularity | `ansible_content_best_practices` or `guidelines://ansible-content-best-practices` |
| **Environment State** | Python, Ansible Core, collections | `ade_environment_info` |
| **Environment Setup** | Provision venv, install deps | `ade_setup_environment` (envName, installRequirements, collections) |
| **Scaffolding** | Collection or playbook project | `ansible_create_collection` / `ansible_create_playbook` |
| **EE Schema Validation** | execution-environment.yml structure | `schema://execution-environment` + `rules://execution-environment` |
| **EE Build** | Container image construction | `define_and_build_execution_env` |
| **Lint Validation** | Static analysis, auto-fix | `ansible_lint` (profile, fix) |
| **Playbook Execution** | Production deployment | `ansible_navigator` |
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.