github-agentic-workflows-tools-ecosystem

name	GitHub Agentic Workflows Tools Ecosystem
description	Comprehensive guide for all available tools including GitHub, file operations, web, bash, playwright, tool capabilities and limitations, integration patterns, custom tool development, security considerations, and usage examples
license	Apache-2.0
version	2.0.1
last_updated	"2026-04-13T00:00:00.000Z"
tags	["github-agentic-workflows","tools","github-tools","file-operations","web-tools","bash","playwright","browser-automation","custom-tools","tool-security","integration-patterns"]

🛠️ GitHub Agentic Workflows Tools Ecosystem

🔴 AI FIRST Quality Principle

Apply the AI FIRST principle: never accept first-pass quality. Minimum 2 iterations. Read all output, improve every section. No shortcuts.

📋 Overview

This skill provides a comprehensive guide to all tools available in GitHub Agentic Workflows. Understanding the tools ecosystem is critical for building effective AI agents that can interact with code, GitHub APIs, browsers, and external services.

What is the Tools Ecosystem?

The Tools Ecosystem is the collection of capabilities that AI agents can invoke to perform actions:

• GitHub Tools: Interact with repositories, issues, PRs, workflows
• File Operations: Read, write, edit files in repositories
• Web Tools: Fetch web pages, make HTTP requests, search the web
• Bash Tools: Execute shell commands, run scripts
• Playwright Tools: Automate browsers, take screenshots, interact with web UIs
• Memory Tools: Store and retrieve knowledge across sessions
• Custom Tools: Develop domain-specific tools via MCP servers

Why Master the Tools Ecosystem?

Effective agents require deep knowledge of available tools:

• ✅ Capability Awareness: Know what's possible
• ✅ Tool Selection: Choose the right tool for each task
• ✅ Efficient Workflows: Combine tools effectively
• ✅ Security: Understand tool limitations and risks
• ✅ Error Handling: Handle tool failures gracefully
• ✅ Extensibility: Build custom tools when needed

---

🐙 GitHub Tools

Repository Operations

View Files and Directories

// github-get_file_contents
await github.getFileContents({
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  path: 'src/index.js',
  ref: 'main',
});

// List directory
await github.getFileContents({
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  path: 'src/',
});

Get Repository Tree

// github-get_repository_tree
await github.getRepositoryTree({
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  tree_sha: 'main',
  recursive: true,
  path_filter: 'src/',
});

Issue Operations

// Create issue
await github.issueWrite({
  method: 'create',
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  title: 'Bug: Application crashes',
  body: 'Description here...',
  labels: ['bug', 'P1-high'],
  assignees: ['developer1'],
});

// Search issues
await github.searchIssues({
  query: 'repo:Hack23/riksdagsmonitor is:open label:bug',
  sort: 'created',
  order: 'desc',
});

Pull Request Operations

// Create PR
await github.createPullRequest({
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  title: 'feat: Add dark mode',
  body: 'Implementation details...',
  head: 'feature/dark-mode',
  base: 'main',
});

// Get PR diff
await github.pullRequestRead({
  method: 'get_diff',
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  pullNumber: 456,
});

// Code review
await github.pullRequestReviewWrite({
  method: 'create',
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  pullNumber: 456,
  body: 'LGTM!',
  event: 'APPROVE',
});

GitHub Actions

// List workflows
await github.actionsList({
  method: 'list_workflows',
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
});

// Trigger workflow
await github.actionsRunTrigger({
  method: 'run_workflow',
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  workflow_id: 'deploy.yml',
  ref: 'main',
  inputs: { environment: 'production' },
});

// Get job logs
await github.getJobLogs({
  owner: 'Hack23',
  repo: 'riksdagsmonitor',
  job_id: 98765432,
  return_content: true,
  tail_lines: 500,
});

---

📁 File Operations

Read Files

// Read entire file
const content = await filesystem.readTextFile({
  path: '/workspace/src/index.js',
});

// Read first 50 lines
const header = await filesystem.readTextFile({
  path: '/workspace/README.md',
  head: 50,
});

// Read last 100 lines
const logs = await filesystem.readTextFile({
  path: '/workspace/app.log',
  tail: 100,
});

// Read multiple files
const files = await filesystem.readMultipleFiles({
  paths: ['file1.js', 'file2.js', 'file3.js'],
});

Write and Edit Files

// Create or overwrite file
await filesystem.writeFile({
  path: '/workspace/output.txt',
  content: 'File content...',
});

// Edit file (line-based)
await filesystem.editFile({
  path: '/workspace/src/index.js',
  edits: [
    {
      oldText: 'const oldVar = 123;',
      newText: 'const newVar = 456;',
    },
  ],
});

Directory Operations

// List directory
const entries = await filesystem.listDirectory({
  path: '/workspace/src',
});

// List with sizes
const entriesWithSizes = await filesystem.listDirectoryWithSizes({
  path: '/workspace/dist',
  sortBy: 'size',
});

// Get directory tree
const tree = await filesystem.directoryTree({
  path: '/workspace',
  excludePatterns: ['node_modules/**', '.git/**'],
});

// Create directory
await filesystem.createDirectory({
  path: '/workspace/new-dir',
});

// Search files
const jsFiles = await filesystem.searchFiles({
  path: '/workspace',
  pattern: '**/*.js',
});

---

🌐 Web Tools

AI-Powered Web Search

// Web search with AI-generated answer
const result = await web.search({
  query: 'What are the latest features in React 19?',
});

// Returns:
// {
//   answer: 'AI-generated answer with citations...',
//   sources: [...]
// }

HTTP Requests (via bash)

# GET request
curl -s https://api.github.com/repos/Hack23/riksdagsmonitor | jq .

# POST request
curl -s -X POST https://api.example.com/endpoint \
  -H "Content-Type: application/json" \
  -d '{"key":"value"}'

# With authentication
curl -s -H "Authorization: Bearer $TOKEN" \
  https://api.example.com/data

---

💻 Bash Tools

Execute Commands

// Synchronous
const result = await bash({
  command: 'npm test',
  mode: 'sync',
  initial_wait: 60,
});

// Async
await bash({
  command: 'npm run build',
  mode: 'async',
  shellId: 'build-session',
});

// Read output
await read_bash({
  shellId: 'build-session',
  delay: 30,
});

// Detached (persistent)
await bash({
  command: 'npm run dev',
  mode: 'async',
  detach: true,
});

Common Patterns

# Code analysis
eslint src/ --format json > report.json
npm audit --json > audit.json

# Git operations
git --no-pager status
git --no-pager diff
git add . && git commit -m "feat: New feature"

# Build and test
npm ci
npm run lint
npm test
npm run build

---

🌐 Playwright Tools

Browser Navigation

// Navigate
await playwright.browserNavigate({
  url: 'https://riksdagsmonitor.pages.dev',
});

// Capture accessibility snapshot
const snapshot = await playwright.browserSnapshot();

Page Interaction

// Click element
await playwright.browserClick({
  ref: '[2]',
  element: 'Button',
});

// Type text
await playwright.browserType({
  ref: '[5]',
  element: 'Search input',
  text: 'query',
  submit: true,
});

// Fill form
await playwright.browserFillForm({
  fields: [
    { ref: '[10]', name: 'username', type: 'textbox', value: 'user' },
    { ref: '[11]', name: 'password', type: 'textbox', value: 'pass' },
  ],
});

Screenshots

// Full page screenshot
await playwright.browserTakeScreenshot({
  filename: 'page.png',
  fullPage: true,
});

// Element screenshot
await playwright.browserTakeScreenshot({
  ref: '[4]',
  element: 'Article',
  filename: 'article.png',
});

---

🧠 Memory Tools

Knowledge Graph

// Create entities
await memory.createEntities({
  entities: [
    {
      name: 'Riksdagsmonitor',
      entityType: 'project',
      observations: ['Static website', 'GitHub Pages'],
    },
  ],
});

// Create relationships
await memory.createRelations({
  relations: [
    {
      from: 'Riksdagsmonitor',
      to: 'GitHub Pages',
      relationType: 'uses',
    },
  ],
});

// Search knowledge
const results = await memory.searchNodes({
  query: 'Swedish Parliament',
});

// Read graph
const graph = await memory.readGraph();

---

🔧 Custom Tool Development

Create MCP Server

// custom-mcp-server.js
import { Server } from '@modelcontextprotocol/sdk/server/index.js';

class CustomMCPServer {
  constructor() {
    this.server = new Server({
      name: 'custom-tools',
      version: '1.0.0',
    }, {
      capabilities: { tools: {} },
    });
    
    this.setupTools();
  }
  
  setupTools() {
    this.server.setRequestHandler('tools/list', async () => ({
      tools: [
        {
          name: 'analyze_code',
          description: 'Analyze code quality',
          inputSchema: {
            type: 'object',
            properties: {
              path: { type: 'string' },
            },
            required: ['path'],
          },
        },
      ],
    }));
    
    this.server.setRequestHandler('tools/call', async (request) => {
      const { name, arguments: args } = request.params;
      
      if (name === 'analyze_code') {
        return this.analyzeCode(args);
      }
      
      throw new Error(`Unknown tool: ${name}`);
    });
  }
  
  async analyzeCode(args) {
    // Implementation
    return {
      content: [
        { type: 'text', text: 'Analysis results...' },
      ],
    };
  }
  
  async start() {
    const transport = new StdioServerTransport();
    await this.server.connect(transport);
  }
}

const server = new CustomMCPServer();
server.start();

Register in Configuration

{
  "mcpServers": {
    "custom-tools": {
      "type": "local",
      "command": "node",
      "args": ["custom-mcp-server.js"],
      "tools": ["*"]
    }
  }
}

---

🔒 Security Considerations

Input Validation

// Validate tool inputs
function validateInput(tool, args) {
  const Ajv = require('ajv');
  const ajv = new Ajv();
  
  const schema = getToolSchema(tool);
  const validate = ajv.compile(schema);
  
  if (!validate(args)) {
    throw new Error('Invalid input');
  }
}

// Sanitize paths
function sanitizePath(path) {
  if (path.includes('..')) {
    throw new Error('Path traversal not allowed');
  }
  return path;
}

Rate Limiting

class RateLimiter {
  constructor(maxRequests, windowMs) {
    this.maxRequests = maxRequests;
    this.windowMs = windowMs;
    this.requests = new Map();
  }
  
  checkLimit(toolName) {
    const now = Date.now();
    const requests = this.requests.get(toolName) || [];
    
    const validRequests = requests.filter(
      time => now - time < this.windowMs
    );
    
    if (validRequests.length >= this.maxRequests) {
      throw new Error('Rate limit exceeded');
    }
    
    validRequests.push(now);
    this.requests.set(toolName, validRequests);
  }
}

Audit Logging

class ToolAuditLogger {
  log(tool, args, result, error = null) {
    const entry = {
      timestamp: new Date().toISOString(),
      tool,
      args: this.sanitizeArgs(args),
      result: error ? null : result,
      error: error ? error.message : null,
    };
    
    console.log(JSON.stringify(entry));
  }
  
  sanitizeArgs(args) {
    const sanitized = { ...args };
    
    for (const key in sanitized) {
      if (key.toLowerCase().includes('token')) {
        sanitized[key] = '[REDACTED]';
      }
    }
    
    return sanitized;
  }
}

---

🔗 Integration Patterns

Sequential Chain

async function sequentialChain(steps) {
  const results = [];
  
  for (const step of steps) {
    const result = await invokeTool(step.tool, step.args);
    results.push(result);
  }
  
  return results;
}

Parallel Execution

async function parallelExecution(tasks) {
  const promises = tasks.map(task =>
    invokeTool(task.tool, task.args)
  );
  
  return await Promise.all(promises);
}

Conditional Execution

async function conditionalExecution(condition, ifTrue, ifFalse) {
  const result = await invokeTool(condition.tool, condition.args);
  
  if (evaluateCondition(result, condition.predicate)) {
    return await invokeTool(ifTrue.tool, ifTrue.args);
  } else if (ifFalse) {
    return await invokeTool(ifFalse.tool, ifFalse.args);
  }
  
  return null;
}

Error Handling

async function retryTool(tool, args, maxRetries = 3) {
  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    try {
      return await invokeTool(tool, args);
    } catch (error) {
      if (attempt === maxRetries) throw error;
      
      const delay = Math.pow(2, attempt) * 1000;
      await new Promise(resolve => setTimeout(resolve, delay));
    }
  }
}

---

📊 Tool Capabilities Matrix

Tool	Read	Write	Search	Execute	Browser
GitHub Issues	✅	✅	✅	❌	❌
GitHub PRs	✅	✅	✅	✅	❌
File System	✅	✅	✅	❌	❌
Bash	✅	✅	❌	✅	❌
Web	✅	❌	✅	❌	❌
Playwright	✅	✅	❌	✅	✅
Memory	✅	✅	✅	❌	❌

---

🎓 Related Skills

• gh-aw-security-architecture: Tool security
• gh-aw-mcp-configuration: Custom tools
• gh-aw-continuous-ai-patterns: Tool orchestration

---

🆕 Tool Scoping & Engine Support (v0.68.1)

Agentic Workflows Introspection Tool

Enable workflow introspection for AI agents with:

tools:
  agentic-workflows: true

This provides tools for:

• status — Show status of workflow files in the repository
• compile — Compile markdown workflows to YAML
• logs — Download and analyze workflow run logs
• audit — Investigate workflow run failures and generate reports
• checks — Classify CI check state for a pull request (returns normalized verdict: success, failed, pending, no_checks, policy_blocked)

Use case: Enable AI agents to analyze GitHub Actions traces and improve workflows based on execution history.

Runtime Configuration

All agentic workflows MUST specify the Node.js runtime version:

runtimes:
  node:
    version: "26"

Runtime configuration properties:

• version: — Runtime version (e.g., "26", "3.12", "latest")
• action-repo: — GitHub Actions setup action (e.g., "actions/setup-node")
• action-version: — Setup action version (e.g., "v4", "v5")
• if: — Optional condition (e.g., "hashFiles('go.mod') != ''")

GitHub Tools Configuration

Scope tool access in frontmatter:

---
tools:
  github:
    toolsets: [issues, labels, pull-requests]  # Specific toolsets
    min-integrity: approved                     # Integrity filtering
---

Available toolsets: all, context, repos, issues, pull-requests, users, projects, actions, security, discussions, stars, notifications, gists (all is a shorthand that enables every GitHub toolset)

Engine-Specific Tool Support

Engine	GitHub Tools	Bash	Playwright	File Ops	MCP
Copilot	✅	✅	✅	✅	✅
Claude	✅	✅	✅	✅	✅
Codex	✅	✅	✅	✅	✅
Gemini	✅	✅	✅	✅	✅

MCP Tool Routing

Connect external tools via MCP servers. The MCP Gateway routes requests from the agent to Docker containers running MCP servers. In this repository, MCP servers are configured via a top-level mcp-servers key in workflow frontmatter, with additional repo-level definitions in .github/copilot-mcp.json:

---
mcp-servers:
  custom-server:
    command: npx
    args: ["-y", "@my/mcp-server"]
tools:
  github:
    toolsets: [issues]
---

📚 References

• GitHub Tools Reference
• Tools Overview
• MCP Specification
• Playwright Docs

---

✅ Remember

• ✅ Scope tools to minimum needed (least privilege)
• ✅ Use toolsets to restrict GitHub API access
• ✅ Use min-integrity for public repo security
• ✅ All engines support the same tool categories
• ✅ MCP enables extensibility with custom servers
• ✅ Validate tool inputs, sanitize paths and commands
• ✅ Rate limit tool invocations
• ✅ Handle errors gracefully
• ✅ Use parallel tool execution when possible
• ✅ Audit log all tool usage

---

Last Updated: 2026-04-02
Version: 2.0.0
License: Apache-2.0

---

🔗 Integration with Riksdagsmonitor agentic workflows

This gh-aw skill is applied by the 11 agentic news workflows in .github/workflows/news-*.md. Their domain contract (analysis-artifact product, gate, article contract) lives in:

• .github/prompts/README.md — module catalogue, import rules, AI-FIRST 2-pass rule.
• analysis/methodologies/ai-driven-analysis-guide.md + analysis/templates/ — 9 core / 14 Tier-C artifacts.
• 05-analysis-gate.md — the single blocking gate before any article content is written.

Upstream gh-aw docs (v0.69.3): abridged · complete · agentic-workflows blog series · source repo · GitHub CLI manual.