agent-builder

Automate Claude agent construction and headless CLI workflows programmatically.

310|45|Updated Feb 22, 2026
One-click install
npx skills add https://github.com/Mathews-Tom/praxis-skills --skill agent-builder-mathews-tom
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-builder
Source: https://github.com/Mathews-Tom/praxis-skills/tree/main/skills/agent-builder
Command: npx skills add https://github.com/Mathews-Tom/praxis-skills --skill agent-builder-mathews-tom

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill eliminates brittle, manual workflows for creating and operating Claude agents by providing repeatable, production-grade patterns for programmatic agent construction and headless automation.

Core Features & Use Cases

  • Agent SDK Patterns: Create ClaudeSDKClient instances, configure ClaudeAgentOptions, and register SDK MCP servers to expose custom tools.
  • Headless CLI Automation: Use claude -p for non-interactive queries, structured JSON output, streaming responses, session resume, and CI/CD integration.
  • Security & Control: Implement hooks for deterministic validation, permission rules for allowed/disallowed tools, budget and turn limits, and explicit error handling.
  • Use Case: Automate a CI pipeline that runs tests, applies safe edits, and re-runs until all tests pass while enforcing tool permissions and session resumption.

Quick Start

Build an agent to run the test suite and apply fixes automatically using the Agent SDK with allowed tools Read, Edit, and Bash in headless mode.

Frequently Asked Questions about agent-builder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate Claude agents programmatically using the Python Agent SDK?

You can automate Claude agents by creating ClaudeSDKClient instances, configuring ClaudeAgentOptions, and registering SDK MCP servers to expose custom tools for programmatic agent construction and headless automation workflows.

Can I run Claude in headless mode for non-interactive queries and structured JSON output?

Yes, you can use the claude -p command for headless CLI automation to execute non-interactive queries, parse structured JSON output, stream responses, and integrate Claude into automated pipelines.

Does the Claude Agent SDK support session management and resumption?

Yes, the Claude Agent SDK supports session management and resumption, allowing you to pause and resume automated agent workflows while maintaining context across headless CLI executions.

How do I enforce tool permissions and deterministic validation with Claude hooks?

You can implement hooks for deterministic validation and configure permission rules for allowed or disallowed tools, set budget and turn limits, and handle errors explicitly to enforce security and control over Claude agents.

What's the best way to integrate Claude into CI/CD automation for applying fixes?

The best approach is using the Agent SDK with headless CLI to automate a CI pipeline that runs tests, applies safe edits with allowed tools like Read and Edit, and re-runs until all tests pass with session resumption.

Do I need raw Anthropic API keys to build Claude agents with the SDK?

No, you do not need raw Anthropic API keys; the Agent SDK and headless CLI workflows use ~/.claude/ authentication, eliminating the need to manage raw API keys directly in your automation code.