### Install and Run Documentation Site Locally

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/contributing.md

Commands to install dependencies, generate rules, and start the local documentation website. Ensure you are in the website directory.

```bash
npm --prefix website ci

```

```bash
npm --prefix website run generate:rules

```

```bash
npm --prefix website start

```

--------------------------------

### CLAUDE.md Structure Template: Setup Commands

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/claude-code-OPINIONS.md

Provides example bash commands for setting up a project, including installing dependencies and copying environment files. This is part of a standard CLAUDE.md structure.

```bash
npm install
cp .env.example .env.local
```

--------------------------------

### Full Agnix Setup Example

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/doc/agnix.txt

Comprehensive setup configuration for the agnix Neovim plugin, including options for command path, filetypes, root markers, autostart, attach callback, LSP settings, and logging.

```lua
require('agnix').setup({
  cmd = nil,              -- auto-detect agnix-lsp binary
  filetypes = { 'markdown', 'json' },
  root_markers = { '.git', '.agnix.toml', 'CLAUDE.md', 'AGENTS.md' },
  autostart = true,
  on_attach = function(client, bufnr)
    -- Custom key mappings, etc.
  end,
  settings = {
    severity = 'Warning',    -- 'Error', 'Warning', 'Info'
    rules = {
      skills = true,
      hooks = true,
      disabled_rules = { 'AS-001' },
    },
    versions = {
      claude_code = '1.0.0',
    },
  },
  log_level = 'warn',
  telescope = { enable = true },
})
```

--------------------------------

### Install and Use pypdf Library

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/agent-skills-OPINIONS.md

Demonstrates how to install the pypdf library using pip and then use it to read a PDF file. This is a good example of providing clear setup instructions.

```python
from pypdf import PdfReader
reader = PdfReader("file.pdf")
```

--------------------------------

### Valid Instructions with Positive Guidance

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/pe-006.md

This example demonstrates prompts that include both negative constraints and positive, constructive guidance. This approach is more effective for guiding AI models.

```markdown
# Rules

Don't use global variables. Instead, pass values as function parameters.
Avoid console.log in production. Use the structured logger instead.
```

--------------------------------

### Install Plugin with Project Scope

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/claude-code-HARD-RULES.md

Example CLI command to install a plugin with the 'project' scope, which uses the .claude/settings.json file for team plugins shared via git.

```bash
claude plugin install formatter@marketplace --scope project
```

--------------------------------

### Install agnix.nvim with vim-plug

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/README.md

Install the agnix Neovim plugin using vim-plug. After adding the Plug line, call the setup function in your init.lua.

```vim
Plug 'agent-sh/agnix'
```

```lua
require('agnix').setup()
```

--------------------------------

### Minimal Agnix Setup

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/doc/agnix.txt

Minimal setup call for the agnix Neovim plugin.

```lua
require('agnix').setup()
```

--------------------------------

### Setup and Run agnix Real-World Validation

Source: https://github.com/agent-sh/agnix/blob/main/docs/REAL-WORLD-TESTING.md

Quick start commands for setting up the Python environment, building the agnix project, and running the validation script against all or a subset of repositories.

```bash
# Setup
python3 -m venv .venv
source .venv/bin/activate
pip install pyyaml

# Build
cargo build --release

# Run against all repos (auto-cleans clones to save disk)
python scripts/real-world-validate.py --parallel 8 --timeout 60

# Run a subset
python scripts/real-world-validate.py --limit 50 --parallel 8
python scripts/real-world-validate.py --category claude-code --parallel 4
python scripts/real-world-validate.py --filter streamlit --parallel 1
```

--------------------------------

### Example AGENTS.md Project Guidelines

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CODEX-REFERENCE.md

An example of an AGENTS.md file outlining project guidelines for code style, testing, and Git conventions.

```markdown
# Project Guidelines

## Code Style
- Use TypeScript strict mode
- Prefer async/await over callbacks
- All functions must have JSDoc comments

## Testing
- Run `npm test` before committing
- Maintain >80% code coverage

## Git Conventions
- Use conventional commits (feat:, fix:, docs:)
- Never force push to main
```

--------------------------------

### Example Structure: Anti-Patterns

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/prompt-engineering-OPINIONS.md

Include 'what-not-to-do' examples to highlight common mistakes and their correct alternatives. This aids in preventing errors.

```yaml
<anti-patterns>
  <anti-pattern>
    ❌ BAD: Creating new file without checking if it exists
    ✅ GOOD: Use Glob to check for existing file, then Edit if found
  </anti-pattern>
</anti-patterns>
```

--------------------------------

### Verify Homebrew Installation

Source: https://github.com/agent-sh/agnix/blob/main/docs/RELEASING.md

Install the agnix CLI using Homebrew and verify the installed version.

```bash
brew install agnix
agnix --version
```

--------------------------------

### Install Dependencies

Source: https://github.com/agent-sh/agnix/blob/main/tests/fixtures/gemini_md/GEMINI.md

Installs all project dependencies using npm.

```bash
npm install
```

--------------------------------

### Install agnix-lsp Manually via npm

Source: https://github.com/agent-sh/agnix/blob/main/editors/zed/README.md

Install the agnix language server globally using npm if automatic download fails. This is the easiest manual installation method.

```bash
# npm (easiest)
npm install -g agnix
```

--------------------------------

### Install agnix via Homebrew

Source: https://github.com/agent-sh/agnix/blob/main/README.md

Install agnix on macOS or Linux using Homebrew by tapping the agnix repository and then installing the package.

```bash
# Homebrew (macOS/Linux)
brew tap agent-sh/agnix && brew install agnix
```

--------------------------------

### Valid Import Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/ref-003.md

This example demonstrates a references file with unique imports, adhering to the REF-003 rule.

```markdown
# Project

@docs/coding-standards.md
@docs/testing-guide.md
```

--------------------------------

### Valid YAML Frontmatter Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cur-003.md

This example demonstrates correctly formatted YAML frontmatter. It is used as a reference for valid configurations.

```markdown
---
description: TypeScript rules
globs: "**/*.ts"
---
# Rules

Use strict mode.

```

--------------------------------

### Minimal Few-Shot Prompting Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/prompt-engineering-OPINIONS.md

Use a minimal number of examples (1-3) for straightforward tasks in few-shot prompting. This example illustrates a simple task with a sequence of actions.

```yaml
<examples>
  <example>
    User: "Fix the bug in auth.py"
    Thought: Need to read the file first to understand the bug
    Action: Read auth.py → Identify issue → Edit to fix → Verify
  </example>
</examples>
```

--------------------------------

### Install Agnix LSP Binary

Source: https://github.com/agent-sh/agnix/blob/main/editors/vscode/README.md

Install the agnix-lsp binary from crates.io or via Homebrew. The binary is automatically downloaded on first use.

```bash
# From crates.io
cargo install agnix-lsp

# Or via Homebrew
brew tap agent-sh/agnix && brew install agnix
```

--------------------------------

### Valid Mode Slug Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/roo-004.md

This example shows a file path with a valid mode slug, demonstrating the correct structure.

```markdown
.roo/rules-architect/general.md
```

--------------------------------

### agnix.setup()

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/doc/agnix.txt

Initializes the Agnix plugin. This function takes an optional `opts` table for configuration, referring to |agnix-setup| for available options.

```APIDOC
## agnix.setup()

### Description
Initializes the plugin. See |agnix-setup| for options.

### Method
Lua Function Call

### Parameters
#### Options Table (`opts`)
- `opts` (table) - Optional - Configuration options for the plugin. Refer to |agnix-setup| for details.
```

--------------------------------

### Example .env File Structure

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/multi-platform-OPINIONS.md

Demonstrates the recommended pattern for an .env.example file, including required and optional variables for different AI models and features.

```bash
# .env.example (commit this)
# Copy to .env and fill in actual values

# Required: Anthropic API key for Claude models
ANTHROPIC_API_KEY=sk-ant-...

# Required: OpenAI API key for GPT models
OPENAI_API_KEY=sk-...

# Optional: Specific model selections
PRIMARY_MODEL=claude-3-5-sonnet-20241022
FALLBACK_MODEL=gpt-4o-mini

# Optional: Feature flags
ENABLE_VOICE=false
ENABLE_MCP_SERVERS=true
```

--------------------------------

### Node.js API: Get Version

Source: https://github.com/agent-sh/agnix/blob/main/npm/README.md

Retrieve the installed version of agnix using the Node.js API.

```javascript
console.log(agnix.version());
```

--------------------------------

### Skill File Structure Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/claude-code-HARD-RULES.md

Shows the directory structure for a skill, including the required SKILL.md entrypoint and optional reference, examples, and scripts directories.

```text
skills/
└── my-skill/
    ├── SKILL.md           # REQUIRED: entrypoint
    ├── reference.md       # Optional: detailed docs
    ├── examples/          # Optional: examples
    └── scripts/           # Optional: executable scripts

```

--------------------------------

### Manual Installation of agnix.nvim

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/README.md

Manually install the agnix Neovim plugin by copying its directory contents to the specified Neovim site-pack path.

```bash
Copy the `editors/neovim/` directory contents to:

~/.local/share/nvim/site/pack/plugins/start/agnix/
```

--------------------------------

### Neovim Editor Setup for Agnix

Source: https://github.com/agent-sh/agnix/blob/main/README.md

Configuration snippet for Neovim to integrate Agnix. Ensure you have the `agnix` plugin installed and configured.

```lua
{ "agent-sh/agnix", config = function() require("agnix").setup() end }
```

--------------------------------

### Multi-Window Workflow Setup (Claude 4.5)

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/PROMPT-ENGINEERING-REFERENCE.md

Steps for managing tasks that span multiple context windows, including setting up frameworks, using structured formats for tests, and creating setup scripts.

```text
1. Use first window to set up framework (tests, scripts)
2. Write tests in structured format (e.g., `tests.json`)
3. Create setup scripts (`init.sh`) for graceful restarts
4. Use git for state tracking across sessions
5. Consider fresh contexts over compaction
```

--------------------------------

### Install agnix LSP Server

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/README.md

Install the agnix LSP server using npm, Cargo, or by downloading a release. This is a prerequisite for the Neovim plugin.

```bash
# npm (easiest)
npm install -g agnix

# Cargo
cargo install agnix-lsp

# Or download from releases
# https://github.com/agent-sh/agnix/releases
```

--------------------------------

### Invalid Skill Name: Leading/Trailing Hyphens

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/agent-skills-HARD-RULES.md

Skill names cannot start or end with a hyphen. These examples demonstrate invalid name formats that will fail validation.

```yaml
name: -pdf-tool  # WILL FAIL VALIDATION
```

```yaml
name: pdf-tool-  # WILL FAIL VALIDATION
```

--------------------------------

### Agent Config Examples: Few-Shot Prompting Best Practices

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/prompt-engineering-HARD-RULES.md

Illustrates good and bad practices for structuring few-shot examples in agent configurations, emphasizing order and balance.

```yaml
✅ GOOD: Place most important examples first (but test variations)
✅ GOOD: Balance example types (not all edge cases, include common cases)
✅ GOOD: Be aware order matters—test different orderings
❌ BAD: Assuming more examples always helps
❌ BAD: Placing all negative examples together at the end
```

--------------------------------

### Valid Asset Path Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cdx-pl-012.md

This JSON snippet shows a valid asset path for the 'logo' field. Relative paths starting with './' are considered valid.

```json
{
  "name": "my-plugin",
  "interface": {"logo": "./assets/logo.png"}
}
```

--------------------------------

### Example MCP Server Implementation

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/multi-platform-OPINIONS.md

This TypeScript example demonstrates how to set up an MCP server for database access, defining tools for actions and resources for data. It uses the Model Context Protocol SDK.

```typescript
// mcp-servers/database/
import { Server } from "@modelcontextprotocol/sdk/server/index.js";

const server = new Server({
  name: "database-mcp-server",
  version: "1.0.0"
}, {
  capabilities: {
    tools: {},
    resources: {},
  }
});

// Define tools (actions)
server.setRequestHandler("tools/list", async () => ({
  tools: [
    {
      name: "query_database",
      description: "Execute SQL query",
      inputSchema: {
        type: "object",
        properties: {
          query: { type: "string" },
          params: { type: "array" }
        }
      }
    }
  ]
}));

// Define resources (data)
server.setRequestHandler("resources/list", async () => ({
  resources: [
    {
      uri: "db://schema/tables",
      name: "Database Schema",
      mimeType: "application/json"
    }
  ]
}));
```

--------------------------------

### Cursor Project-Specific Rules

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/multi-platform-OPINIONS.md

Example of custom rules for Cursor to guide the AI on project-specific anti-patterns and constraints. This helps the AI learn unique project requirements.

```markdown
## This Project's Gotchas
- Never use `Date.now()` directly - use our `getCurrentTimestamp()` utility
- Database queries must use prepared statements (prevent SQL injection)
- All API responses must match our standard error format (see ErrorResponse type)
```

--------------------------------

### Integration Test Script for Claude

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/claude-code-OPINIONS.md

An example shell script demonstrating how to use Claude for integration testing, including project setup, command execution, and file verification.

```bash
set -e

echo "Setting up test environment..."
mkdir -p /tmp/test-project
cd /tmp/test-project
git init

echo "Loading plugins..."
claude --plugin-dir ../plugins \
       --session-id integration-test \
       --command "/create-component Button" \
       --exit-after-command

echo "Verifying files created..."
[ -f "src/components/Button.tsx" ] || exit 1
[ -f "src/components/Button.test.tsx" ] || exit 1

echo "Running tests..."
npm test Button

echo "Verifying lint..."
npm run lint src/components/Button.tsx

echo "✅ Integration test passed"
```

--------------------------------

### Resource Picker UI Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-OPINIONS.md

Demonstrates a user interface for resource selection, featuring a hierarchical view with checkboxes for selection and options for bulk actions.

```text
📁 Resources
├─ 📁 Project Files
│  ├─ ☑ README.md (18 KB)
│  ├─ ☐ CONTRIBUTING.md (5 KB)
│  └─ ☑ src/main.py (42 KB)
├─ 📁 Documentation
│  └─ ☐ API.md (103 KB)
└─ 🔗 External
   └─ ☑ https://api.example.com/schema

[Select All] [Clear] [Add Selected]
```

--------------------------------

### Example .agnix.toml Configuration

Source: https://github.com/agent-sh/agnix/blob/main/editors/zed/README.md

Configure agnix linting rules and severity for your project. This example sets the target and disables a specific rule.

```toml
target = "claude-code"
severity = "warning"

[rules]
disabled_rules = ["AS-001"]
```

--------------------------------

### Effective System Prompt Example for Constraints

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/PROMPT-ENGINEERING-REFERENCE.md

Illustrates how to provide context for negative constraints, explaining the 'why' behind the rule for better model adherence.

```text
Your response will be read aloud by a text-to-speech engine, so never use
ellipses since the text-to-speech engine will not know how to pronounce them.
```

--------------------------------

### End-to-End Integration Test for MCP Client

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-OPINIONS.md

An asynchronous Python example demonstrating an end-to-end integration test for an MCP client. It starts a server as a subprocess and interacts with it using ClientSession.

```python
import asyncio
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client

async def test_end_to_end():
    # Start server as subprocess
    server_params = StdioServerParameters(
        command="python",
        args=["server.py"]
    )

    async with stdio_client(server_params) as (read, write):
        async with ClientSession(read, write) as session:
            # Initialize
            await session.initialize()

            # List tools
            tools = await session.list_tools()
            assert len(tools) > 0

            # Call tool
            result = await session.call_tool(
                "get_weather",
                {"location": "New York"}
            )
            assert result is not None
```

--------------------------------

### Full Configuration Reference

Source: https://github.com/agent-sh/agnix/blob/main/docs/CONFIGURATION.md

A comprehensive TOML configuration example covering severity, target, tools, file exclusions, rule toggles, and version specifications.

```toml
severity = "Warning"  # Warning, Error, Info
target = "Generic"    # Deprecated: Generic, ClaudeCode, Cursor, Codex

# Multi-tool support (overrides target)
tools = ["claude-code", "cursor", "github-copilot"]  # Valid: claude-code, cursor, codex, copilot, github-copilot, generic

exclude = [
  "node_modules/**",
  ".git/**",
  "target/**",
]

[rules]
# Category toggles - all default to true
skills = true              # AS-*, CC-SK-* rules
hooks = true               # CC-HK-* rules
agents = true              # CC-AG-* rules
copilot = true             # COP-* rules
cursor = true              # CUR-* rules
memory = true              # CC-MEM-* rules
plugins = true             # CC-PL-* rules
mcp = true                 # MCP-* rules
prompt_engineering = true  # PE-* rules
xml = true                 # XML-* rules
imports = true             # REF-* rules
cross_platform = true      # XP-* rules
agents_md = true           # AGM-* rules

# Disable specific rules by ID
disabled_rules = ["CC-MEM-006", "PE-003"]

# Version-aware validation (optional)
[tool_versions]
# claude_code = "1.0.0"
# cursor = "0.45.0"

[spec_revisions]
# mcp_protocol = "2025-11-25"

# File inclusion/exclusion for non-standard agent files
[files]
# Validate as CLAUDE.md-like memory/instruction files
# include_as_memory = ["docs/ai-rules/*.md", "custom/INSTRUCTIONS.md"]

# Validate as generic markdown (XML, imports, cross-platform rules)
# include_as_generic = ["internal/*.md"]

# Exclude from validation entirely (even built-in file types)
# exclude = ["vendor/**", "generated/**"]

# Per-file rule suppression (see "Per-file rule overrides" below).
# Each [[overrides]] block disables `disabled_rules` for files matching
# any pattern in `paths`. Multiple blocks stack (set union); ordering
# does not matter.
# [[overrides]]
# paths = ["CLAUDE.md", "AGENTS.md"]
# disabled_rules = ["CC-MEM-005"]

```

--------------------------------

### Prompting Decision Framework: Zero-Shot vs Few-Shot

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/PROMPT-ENGINEERING-REFERENCE.md

A flowchart illustrating the decision process for choosing between zero-shot and few-shot prompting, starting with zero-shot and progressively adding examples if needed.

```text
Start with Zero-Shot
    |
    v
Does it work? --Yes--> Done
    |
    No
    v
Add 2-3 Few-Shot Examples
    |
    v
Does it work? --Yes--> Done
    |
    No
    v
Add more examples OR consider fine-tuning
```

--------------------------------

### Invalid YAML Frontmatter: Missing Delimiters

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/agent-skills-HARD-RULES.md

YAML frontmatter in markdown files must be correctly delimited by '---' on both the start and end. This example shows missing closing delimiters.

```markdown
name: my-skill
description: Does something
---

# Instructions
```

```markdown
---
name: my-skill
description: Does something

# Instructions (missing closing ---)
```

--------------------------------

### Valid copilot-setup-steps Job Configuration

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cop-018.md

This example demonstrates a valid job configuration that adheres to the COP-018 rule. Correctly naming the job 'copilot-setup-steps' resolves the issue.

```yaml
jobs:
  copilot-setup-steps:
    runs-on: ubuntu-latest
    steps:
      - run: echo setup
```

--------------------------------

### Valid Agent Skills File Name

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/as-005.md

This example demonstrates a correctly formatted agent skills file name that does not start or end with a hyphen, satisfying rule AS-005. This configuration is compliant.

```markdown
---
name: build-project
description: Use when building the project
---
Run the build script.
```

--------------------------------

### Effective System Prompt Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/PROMPT-ENGINEERING-REFERENCE.md

Contrast less effective, vague instructions with more effective, explicit instructions for creating an analytics dashboard.

```text
Create an analytics dashboard. Include as many relevant features and
interactions as possible. Go beyond the basics to create a fully-featured
implementation.
```

--------------------------------

### Invalid Agent Skills File Name

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/as-005.md

This example shows an agent skills file name that starts and ends with a hyphen, which violates rule AS-005. This configuration would be flagged by the linter.

```markdown
---
name: -build-project-
description: Use when building the project
---
Run the build script.
```

--------------------------------

### Basic Codex SDK Usage

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CODEX-REFERENCE.md

A fundamental example of using the Codex TypeScript SDK to start a thread, run a task, and access the final response. It demonstrates the core workflow for interacting with Codex via code.

```typescript
import { Codex } from "@openai/codex-sdk";

const codex = new Codex();
const thread = codex.startThread({
  workingDirectory: process.cwd()
});

const turn = await thread.run("Find and fix bugs");
console.log(turn.finalResponse);
```

--------------------------------

### Slash Command Prompt Discovery Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-OPINIONS.md

Shows an example of a slash command interface for discovering and invoking prompts, with a search input and a list of available commands.

```text
/ [Type to search prompts]

📝 /draft_email - Create a professional email
🔍 /analyze_data - Perform data analysis
✈️ /plan_vacation - Plan a trip itinerary
🐛 /debug_code - Debug code issues
```

--------------------------------

### Invalid agents.md file exceeding size limit

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cdx-ag-004.md

This TOML snippet represents an invalid agents.md file that triggers the CDX-AG-004 rule due to exceeding the size limit. This example is for demonstration purposes and requires no specific setup.

```toml
agents.md
```

--------------------------------

### agnix.start()

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/doc/agnix.txt

Starts the LSP client for the current Neovim buffer. This function is used to initiate the language server process.

```APIDOC
## agnix.start()

### Description
Start the LSP client for the current buffer.

### Method
Lua Function Call
```

--------------------------------

### Basic Agnix GitHub Action Setup

Source: https://github.com/agent-sh/agnix/blob/main/docs/CONFIGURATION.md

A minimal configuration for the Agnix GitHub Action. This sets up the action to run with default parameters.

```yaml
- uses: agent-sh/agnix@v0
```

--------------------------------

### Invalid Claude Skills Configuration (Too Many Injections)

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cc-sk-009.md

This example shows an invalid Claude Skills configuration that triggers CC-SK-009 due to an excessive number of injections. Ensure the number of injections is appropriate for the skill's setup.

```markdown
---
name: template-skill
description: Use when applying a code template
---
Use !`config.json` and !`env.json` and !`secrets.json` and !`overrides.json` for setup.
```

--------------------------------

### Verify npm Installation

Source: https://github.com/agent-sh/agnix/blob/main/docs/RELEASING.md

Install the agnix CLI using npm and verify the installed version.

```bash
npm install -g @agnix/cli
agnix --version
```

--------------------------------

### Build agnix-lsp from Source

Source: https://github.com/agent-sh/agnix/blob/main/crates/agnix-lsp/README.md

Build the agnix-lsp binary in release mode from the workspace root.

```bash
cargo build --release -p agnix-lsp
```

--------------------------------

### Agent Config Examples: Instruction Length vs. Effectiveness

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/prompt-engineering-HARD-RULES.md

Shows how to balance instruction length and detail, emphasizing task-relevant information over extraneous details.

```yaml
❌ BAD: Long backstory about why the agent exists
❌ BAD: Philosophical explanations about AI behavior
✅ GOOD: Direct, task-relevant instructions
✅ GOOD: Context that affects decision-making
✅ GOOD: Specific examples of desired behavior
```

--------------------------------

### Valid Configuration Using Environment Variables

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/xp-003.md

This example demonstrates a valid configuration that avoids hard-coded platform paths by using environment variables or relative paths. It promotes cross-platform compatibility.

```markdown
# Configuration

Use environment variables for all platform-specific settings.
See ./src/config.ts for defaults.
```

--------------------------------

### Verify Cargo Installation

Source: https://github.com/agent-sh/agnix/blob/main/docs/RELEASING.md

Install the agnix CLI using Cargo and verify the installed version.

```bash
cargo install agnix
agnix --version
```

--------------------------------

### Install Agnix LSP Binary with npm or Cargo

Source: https://github.com/agent-sh/agnix/blob/main/docs/EDITOR-SETUP.md

Use npm to install the Agnix LSP globally or Cargo to install the `agnix-lsp` binary. This is useful for manual installation or when automatic downloads fail.

```bash
# npm
npm install -g agnix
```

```bash
# Cargo
cargo install agnix-lsp
```

--------------------------------

### Example .agnix.toml Configuration

Source: https://github.com/agent-sh/agnix/blob/main/editors/vscode/README.md

Use a .agnix.toml file in your workspace root for team-shared configuration. This example sets the target tool and disables specific rules.

```toml
target = "ClaudeCode"

[rules]
disabled_rules = ["PE-003"]
```

--------------------------------

### Verify agnix installation

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/installation.md

Check if agnix is installed correctly and display its version. This command is useful after any installation method.

```bash
agnix --version
```

--------------------------------

### Valid Profile Configuration

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cdx-app-003.md

This example demonstrates a valid profile configuration. The 'profile' setting is correctly set to a string value.

```toml
profile = "default"
```

--------------------------------

### Tool Invocation UI Flow Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-OPINIONS.md

Illustrates a recommended user interface flow for tool invocation, showing pre-execution, during execution, and post-execution states.

```text
🔧 About to call: send_email
   To: team@example.com
   Subject: "Weekly update"
   Body: [Preview...]

   [Approve] [Modify] [Cancel]

⏳ Sending email...

✅ Email sent successfully at 10:30 AM
```

--------------------------------

### Optional Skill Directory Structure

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/agent-skills-OPINIONS.md

This example outlines a recommended directory structure for organizing skill-related files, including executable code, additional documentation, and static resources.

```tree
skill-name/
├── SKILL.md
├── scripts/          # Executable code
├── references/       # Additional documentation
└── assets/           # Static resources
```

--------------------------------

### Install agnix via npm

Source: https://github.com/agent-sh/agnix/blob/main/README.md

Install the agnix package globally using npm for cross-platform compatibility. This is the recommended installation method.

```bash
# npm (recommended, all platforms)
npm install -g agnix
```

--------------------------------

### Using Symbolic Links for Configuration Management

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/multi-platform-OPINIONS.md

Demonstrates how to use symbolic links to manage scattered configurations, specifically for the Cursor tool. This helps centralize settings within a project.

```bash
ln -s .platform/cursor/.cursor/rules .cursor/rules
```

--------------------------------

### Check Agnix Installation

Source: https://github.com/agent-sh/agnix/blob/main/skills/agnix/SKILL.md

Verify if the agnix CLI tool is installed by checking its version. If not found, instructions are provided to install it using cargo.

```bash
agnix --version
```

```bash
cargo install agnix-cli
```

--------------------------------

### Commit Message Generation Example

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/agent-skills-OPINIONS.md

Provide input/output examples for skills that generate structured text like commit messages. This example demonstrates the desired format.

```markdown
## Commit message format

Generate commit messages following these examples:

**Example 1:**
Input: Added user authentication with JWT tokens
Output:
```
feat(auth): implement JWT-based authentication

Add login endpoint and token validation middleware
```

Follow this style: type(scope): brief description, then detailed explanation.
```

--------------------------------

### Agent Config Examples: Instruction Separation and Structure

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/prompt-engineering-HARD-RULES.md

Illustrates effective prompt structuring using separators to distinguish instructions, context, and input data.

```yaml
✅ GOOD: Use clear sections
---
# Instructions
Edit existing files. Only create new files when explicitly requested.

# Context
You are working in a git repository at /path/to/repo

# Output Format
Respond with file paths and explanations.
---

❌ BAD: Mixed instructions and context in a single paragraph
```

--------------------------------

### Valid Prompt Example: Specific Instructions

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/pe-005.md

This example demonstrates a prompt with specific and actionable instructions. It includes details like formatting requirements (2-space indentation) and error code inclusion, making it more effective.

```markdown
# Rules

Format all output as JSON with 2-space indentation.
Always include error codes in responses.

```

--------------------------------

### Install VS Code Extension from Source

Source: https://github.com/agent-sh/agnix/blob/main/docs/EDITOR-SETUP.md

Steps to install the VS Code extension for agnix from its source code. This involves installing dependencies, compiling, and packaging the extension.

```bash
cd editors/vscode
npm install
npm run compile
npm run package
code --install-extension agnix-*.vsix
```

--------------------------------

### Example tools/list Response

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-HARD-RULES.md

A sample response from the 'tools/list' method, containing an array of tool definitions. Each tool object must include at least its name, description, and inputSchema.

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "tools": [
      {
        "name": "get_weather",
        "description": "Get weather data",
        "inputSchema": { "type": "object", "properties": { "location": { "type": "string" } }, "required": ["location"] }
      }
    ]
  }
}
```

--------------------------------

### Correct Pattern: Initialize First

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-HARD-RULES.md

Shows the correct client behavior of sending an `initialize` request before other methods like `tools/list`.

```json
// ✅ CORRECT: Initialize first
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "initialize",
  "params": { "protocolVersion": "2025-11-25", ... }
}
// Wait for response, send initialized notification, then send tools/list
```

--------------------------------

### Install Dependencies with pnpm

Source: https://github.com/agent-sh/agnix/blob/main/tests/fixtures/cross_platform/conflicting-commands/AGENTS.md

Installs project dependencies using pnpm.

```bash
pnpm install
```

--------------------------------

### Valid Prompt File with Content

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cop-013.md

This example demonstrates a valid prompt file that adheres to the COP-013 rule. It includes a description and a non-empty body, providing necessary context for GitHub Copilot.

```markdown
---
description: Refactor selected code
---
Refactor the selected code while preserving behavior.
```

--------------------------------

### Install Agnix CLI

Source: https://github.com/agent-sh/agnix/blob/main/crates/agnix-cli/README.md

Install the agnix CLI using cargo.

```bash
cargo install agnix-cli
```

--------------------------------

### MCP Server CLI Wizard

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CLAUDE-CODE-REFERENCE.md

Initiate the MCP server setup process using the `claude mcp add` command.

```bash
claude mcp add
```

--------------------------------

### Neovim Lazy.nvim Installation

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/editor-integration.md

Install the agnix.nvim plugin using the lazy.nvim package manager.

```lua
{ "agent-sh/agnix.nvim" }
```

--------------------------------

### TypeScript MCP Server Installation

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CLAUDE-CODE-REFERENCE.md

Install the MCP TypeScript SDK using npm.

```bash
npm install @modelcontextprotocol/sdk
```

--------------------------------

### Example tools/call Request

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-HARD-RULES.md

Demonstrates a request to the 'tools/call' method, specifying the tool name and its arguments. The server must validate the provided arguments against the tool's inputSchema.

```json
{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "get_weather",
    "arguments": {
      "location": "New York"
    }
  }
}
```

--------------------------------

### Python MCP Server Installation

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CLAUDE-CODE-REFERENCE.md

Install the MCP Python SDK using pip.

```bash
pip install mcp
```

--------------------------------

### Example Version Mismatch Error

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-HARD-RULES.md

An example of an initialization error where the protocol version is mismatched.

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "error": {
    "code": -32602,
    "message": "Unsupported protocol version",
    "data": {
      "supported": ["2024-11-05"],
      "requested": "1.0.0"
    }
  }
}
```

--------------------------------

### Agnix Configuration File Example

Source: https://github.com/agent-sh/agnix/blob/main/SPEC.md

Shows the structure of the .agnix.toml configuration file, including settings for severity, target, locale, and rule customization. This file allows for persistent configuration of Agnix behavior.

```toml
severity = "Warning"
target = "Generic"  # Options: Generic, ClaudeCode, Cursor, Codex
locale = "en"       # Options: en, es, zh-CN
tools = ["claude-code", "cursor"]  # Preferred over target

[rules]

```

--------------------------------

### Valid Frontmatter Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/cop-002.md

This example demonstrates the correct format for scoped instructions with valid frontmatter, ensuring that GitHub Copilot correctly applies the specified instructions to the target files. The `applyTo` field specifies the file pattern.

```markdown
---
applyTo: "**/*.ts"
---
# TypeScript Instructions

Use strict mode and explicit types.

```

--------------------------------

### Install Agnix with vim-plug

Source: https://github.com/agent-sh/agnix/blob/main/editors/neovim/doc/agnix.txt

Configuration for installing the agnix plugin using the vim-plug package manager.

```vim
Plug 'agent-sh/agnix'

" In your init.lua or after/plugin:
lua require('agnix').setup()
```

--------------------------------

### Example prompts/get Request

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/mcp-HARD-RULES.md

A JSON-RPC request to retrieve a specific prompt by its name and provide necessary arguments for its execution.

```json
{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "prompts/get",
  "params": {
    "name": "code_review",
    "arguments": {
      "code": "def hello():\n    print('world')"
    }
  }
}
```

--------------------------------

### Install agnix-lsp with Cargo

Source: https://github.com/agent-sh/agnix/blob/main/crates/agnix-lsp/README.md

Install the agnix-lsp binary using Cargo, the Rust package manager.

```bash
cargo install agnix-lsp
```

--------------------------------

### Install Claude Agent SDK (TypeScript)

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CLAUDE-CODE-REFERENCE.md

Install the Claude Agent SDK for TypeScript using npm.

```bash
npm install @anthropic-ai/claude-agent-sdk
```

--------------------------------

### Validate .env.example Completeness

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/standards/multi-platform-OPINIONS.md

Ensures that all required environment variables are present in the .env.example file. This is crucial for setting up new environments correctly.

```bash
required_keys=("ANTHROPIC_API_KEY" "OPENAI_API_KEY")
for key in "${required_keys[@]}"; do
  grep -q "$key" .env.example || exit 1
done
```

--------------------------------

### Valid Configuration Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/xp-001.md

This configuration is valid as it adheres to generic project guidelines without platform-specific features.

```markdown
# Project Guidelines

Follow the coding style guide.

## Commands
- npm run build
- npm run test
```

--------------------------------

### Install Claude Agent SDK (Python)

Source: https://github.com/agent-sh/agnix/blob/main/knowledge-base/agent-docs/CLAUDE-CODE-REFERENCE.md

Install the Claude Agent SDK for Python using pip.

```bash
pip install claude-agent-sdk
```

--------------------------------

### Build Project

Source: https://github.com/agent-sh/agnix/blob/main/tests/fixtures/gemini_md/GEMINI.md

Builds the project for production deployment.

```bash
npm run build
```

--------------------------------

### Valid agents.md Example (With Project Context)

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/agm-004.md

This example shows a correctly formatted agents.md file that includes a project context section, satisfying the AGM-004 rule. It clearly defines the project and its purpose.

```markdown
# Project

This is a web application for task management built with React and Node.js.

## Build Commands

Run npm install and npm build.

```

--------------------------------

### Invalid JSON Example

Source: https://github.com/agent-sh/agnix/blob/main/website/docs/rules/generated/gm-009.md

This snippet shows an example of invalid JSON that would trigger the GM-009 rule.

```json
{ invalid json }
```

--------------------------------

### Agnix CLI Reference Examples

Source: https://github.com/agent-sh/agnix/blob/main/plugin/skills/agnix/SKILL.md

Provides examples of common agnix CLI commands and their functionalities, including validation, auto-fixing, strict mode, targeting specific AI tools, watch mode, and output formatting.

```bash
agnix .
```

```bash
agnix --fix .
```

```bash
agnix --strict .
```

```bash
agnix --target claude-code .
```

```bash
agnix --target cursor .
```

```bash
agnix --watch .
```

```bash
agnix --format json .
```

```bash
agnix --format sarif .
```