agent-builder
Agent Builder guides you through creating AI agents using the Claude Agent SDK and headless mode—no raw API keys needed. Build custom tools with MCP servers, control execution with hooks, and deploy agents in Python or via CLI commands. Covers SDK patterns, security practices, and production workflows.
Agent Builder helps you create and deploy AI agents programmatically using Claude's SDK and headless CLI mode.
AI-generated summary based on this skill's SKILL.md
Install
Mathews-Tom/armory/agent-builder · repository language: Python
git clone https://github.com/Mathews-Tom/armory
cp -r armory/skills/agent-builder ~/.claude/skills/agent-builderFrequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
How do I build an AI agent with Claude SDK?
Agent Builder helps you construct AI agents using the Claude Agent SDK, which provides a higher-level abstraction than raw API calls. Start by installing the SDK, then define your agent's tools and capabilities using MCP servers or custom tool definitions. The SDK handles session management, permission rules, and execution hooks automatically, letting you focus on agent logic rather than low-level API details.
What is Claude headless mode and how do I use it?
Agent Builder's headless mode (invoked via `claude -p` CLI commands) lets you run Claude Code and agents without a GUI—ideal for automation, scripting, and CI/CD pipelines. In headless mode, Agent Builder manages authentication internally, so you don't expose raw API keys. Use it for programmatic agent creation, batch processing, and integration into application workflows.
How do I create custom tools and MCP servers for agents?
Agent Builder supports custom tool creation through MCP (Model Context Protocol) servers. Define your tools as MCP endpoints, then register them with your agent configuration. Agent Builder handles tool invocation, parameter validation, and response formatting. This approach lets you extend agent capabilities beyond built-in functions while maintaining security and permission controls.
Can I integrate Agent Builder agents into CI/CD pipelines?
Yes. Agent Builder's headless mode and Python SDK make CI/CD integration straightforward. Deploy agents as containerized services or invoke them via CLI in pipeline steps. Agent Builder manages session state and execution hooks, so your pipeline can trigger agents, pass parameters, and capture results without manual API management or key exposure.
What are the key differences between Agent SDK and raw API patterns?
Agent Builder's Agent SDK abstracts authentication, session management, and permission enforcement—you don't handle raw API keys or low-level HTTP calls. The SDK provides hooks for custom logic, built-in tool registration, and MCP server integration. Raw API patterns require manual state tracking and security handling, making the SDK the recommended approach for production agent deployments.
How does Agent Builder handle authentication and permissions?
Agent Builder manages authentication internally in headless mode, eliminating the need to pass raw API keys in scripts or CI/CD environments. Permission rules are enforced at the agent level—define what tools and resources your agent can access during setup. The SDK validates permissions on each execution, ensuring secure, auditable agent behavior across all deployment contexts.
SKILL.md
rendered from the published skill — quoted content, verbatim
Agent Builder - Claude SDK & CLI Expert
Triggers:
- "build agent", "create agent", "agent sdk", "claude sdk"
- "headless mode", "programmatic", "cli command"
- "claude -p", "query programmatically"
- "custom tools", "mcp server", "sdk mcp"
- "claude code api", "agent options"
Prefer Claude Code Headless Mode
This skill covers Claude Code's headless mode and the Agent SDK — not the raw Anthropic API.
Preferred: Claude Code CLI headless mode (claude -p)
Preferred: Claude Agent SDK (wraps Claude Code CLI)
Avoid: Raw Anthropic API (anthropic.Anthropic())
Avoid: Direct API calls with API keys
Why this matters:
- The Agent SDK wraps the Claude Code CLI — it does not use the raw Anthropic API
(truncated - see the full file via the links below)
Read as markdown · JSON record · Browse the source repository
File tree — 10 files
skills/agent-builder/SKILL.md
skills/agent-builder/evals/cases.yaml
skills/agent-builder/references/api-vs-sdk.md
skills/agent-builder/references/best-practices.md
skills/agent-builder/references/cli-reference.md
skills/agent-builder/references/example-agent.md
skills/agent-builder/references/headless-cli.md
skills/agent-builder/references/output-format.md
skills/agent-builder/references/sdk-python.md
skills/agent-builder/references/use-cases.md