Automate pull request reviews by using the anthropics/claude-code-action@beta GitHub Action. The workflow can be triggered on PR events or when a user mentions @claude in an issue comment.

Required Setup: You must add ANTHROPIC_API_KEY to your repository secrets: Settings → Secrets and variables → Actions → New repository secret.

name: PR - Claude Code Review
on:
  pull_request:
    types: [opened, synchronize, reopened]
  issue_comment:
    types: [created]

jobs:
  review:
    if: |
      github.event_name == 'pull_request' ||
      (github.event_name == 'issue_comment' &&
       github.event.issue.pull_request &&
       contains(github.event.comment.body, '@claude'))
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - uses: anthropics/claude-code-action@beta
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          model: claude-opus-4-5-20251101
          prompt: |
            Review this PR using .claude/agents/code-reviewer.md standards.
            Run `git diff origin/main...HEAD` to see changes.

Skills are markdown files located at .claude/skills/{skill-name}/SKILL.md that teach Claude project-specific patterns. Claude uses the description field to decide when to apply the skill.

SKILL.md Format

---
name: skill-name
description: What this skill does and when to use it.
allowed-tools: Read, Grep, Bash(npm:*)
model: claude-sonnet-4-20250514
---

# Skill Title

## When to Use
- Trigger condition

## Core Patterns
```typescript
// Example

### Frontmatter Fields
- `name`: (Required) Lowercase, numbers, and hyphens only. Must match directory name.
- `description`: (Required) Max 1024 chars. Used for semantic matching.
- `allowed-tools`: (Optional) Comma-separated list of tools (e.g., `Read, Grep`).
- `model`: (Optional) Specific model (e.g., `sonnet`, `opus`, `haiku`).

name: skill-name description: What this skill does and when to use it. Include keywords users would mention. allowed-tools: Read, Grep, Glob model: claude-sonnet-4-20250514

Skill Title

When to Use

• Trigger condition 1

Core Patterns

Pattern Name

// Example code

Language Server Protocol (LSP) provides Claude with real-time diagnostics, type information, code navigation, and completions. Enable it via plugins in .claude/settings.json.

Available Plugins

• typescript-lsp@claude-plugins-official: Requires npm install -g typescript-language-server typescript.
• pyright-lsp@claude-plugins-official: Requires pip install pyright.
• rust-lsp@claude-plugins-official: Requires rustup component add rust-analyzer.

Enabling Plugins

{
  "enabledPlugins": {
    "typescript-lsp@claude-plugins-official": true,
    "pyright-lsp@claude-plugins-official": true
  }
}

Custom LSP Configuration

Create a .lsp.json file for advanced setups:

{
  "typescript": {
    "command": "typescript-language-server",
    "args": ["--stdio"],
    "extensionToLanguage": {
      ".ts": "typescript",
      ".tsx": "typescriptreact"
    }
  }
}

Troubleshooting

• Verify binary installation: which <binary-name>
• Enable debug logging: claude --enable-lsp-logging
• Check plugin status: claude /plugin

The skill evaluation system uses the UserPromptSubmit hook to analyze prompts and suggest relevant domain knowledge (Skills).

Setup Steps

1. Copy hooks to your project: cp -r .claude/hooks/ your-project/.claude/hooks/
2. Register the hook in .claude/settings.json:

{
  "hooks": {
    "UserPromptSubmit": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "\"$CLAUDE_PROJECT_DIR\".claude/hooks/skill-eval.sh",
            "timeout": 5
          }
        ]
      }
    ]
  }
}

3. Define triggers in .claude/hooks/skill-rules.json using keywords, keywordPatterns, pathPatterns, intentPatterns, or directoryMatch.

Skill Rules Configuration Example

{
  "testing-patterns": {
    "description": "Jest testing patterns and TDD workflow",
    "priority": 9,
    "triggers": {
      "keywords": ["test", "jest", "spec"],
      "pathPatterns": ["**/*.test.ts"]
    }
  }
}

Agents are specialized assistants with focused purposes and custom system prompts. They are stored in .claude/agents/{agent-name}.md.

Agent Format

---
name: agent-name
description: When/why to use (max 1024 chars).
model: sonnet
---

# Agent System Prompt

Your task is to...

## Checklist
- [ ] Task 1

Configuration Fields

• name: (Required) Lowercase with hyphens.
• description: (Required) Purpose and usage instructions.
• model: (Optional) sonnet, opus, or haiku.
• tools: (Optional) Comma-separated tool list.

---
name: code-reviewer
description: Reviews code for quality, security, and conventions. Use after writing or modifying code.
model: opus
---

# Agent System Prompt

You are a senior code reviewer...

## Your Process
1. Run `git diff` to see changes
2. Apply review checklist
3. Provide feedback

## Checklist
- [ ] No TypeScript `any`
- [ ] Error handling present
- [ ] Tests included

The .claude/settings.json file manages hooks, environment variables, and permissions.

Hook Events

• PreToolUse: Fires before tool execution. Use this to block edits on sensitive branches or validate commands.
• PostToolUse: Fires after a tool completes. Use this for auto-formatting, running tests, or linting.
• UserPromptSubmit: Fires when a user submits a prompt. Use this to add context or suggest skills.
• Stop: Fires when the agent finishes. Use this to decide if Claude should continue.

Hook Response Format

Hooks should return a JSON object with these keys:

• block: (boolean) Block the action (PreToolUse only).
• message: (string) Message to show the user.
• feedback: (string) Non-blocking feedback.
• suppressOutput: (boolean) Hide command output.
• continue: (boolean) Whether to continue.

{
  "block": true,
  "message": "Reason",
  "feedback": "Info",
  "suppressOutput": true,
  "continue": false
}