claude-agent-sdk

Name: claude-agent-sdk
Author: jackspace

15stars7forks

How to install claude-agent-sdk?

jackspace/claudeskillz/claude-agent-sdk

$npx -y skills add jackspace/ClaudeSkillz --skill claude-agent-sdk

Installs into the current project.

Summary

This skill provides comprehensive knowledge for working with the Anthropic Claude Agent SDK. It should be used when building autonomous AI agents, creating multi-step reasoning workflows, orchestrating specialized subagents, integrating custom tools and MCP servers, or implementi

Files · 1

View on GitHub

SKILL.md

1	# Claude Agent SDK
2
3	Status: Production Ready
4	Last Updated: 2025-10-25
5	Dependencies: @anthropic-ai/claude-agent-sdk, zod
6	Latest Versions: @anthropic-ai/claude-agent-sdk@0.1.0+, zod@3.23.0+
7
8	---
9
10	## Quick Start (5 Minutes)
11
12	### 1. Install SDK
13
14	```bash
15	npm install @anthropic-ai/claude-agent-sdk zod
16	```
17
18	Why these packages:
19	- `@anthropic-ai/claude-agent-sdk` - Main Agent SDK
20	- `zod` - Type-safe schema validation for tools
21
22	### 2. Set API Key
23
24	```bash
25	export ANTHROPIC_API_KEY="sk-ant-..."
26	```
27
28	CRITICAL:
29	- API key required for all agent operations
30	- Never commit API keys to version control
31	- Use environment variables
32
33	### 3. Basic Query
34
35	```typescript
36	import { query } from "@anthropic-ai/claude-agent-sdk";
37
38	const response = query({
39	prompt: "Analyze the codebase and suggest improvements",
40	options: {
41	model: "claude-sonnet-4-5",
42	workingDirectory: process.cwd(),
43	allowedTools: ["Read", "Grep", "Glob"]
44	}
45	});
46
47	for await (const message of response) {
48	if (message.type === 'assistant') {
49	console.log(message.content);
50	}
51	}
52	```
53
54	---
55
56	## The Complete Claude Agent SDK Reference
57
58	## Table of Contents
59
60	1. [Core Query API](#core-query-api)
61	2. [Tool Integration](#tool-integration-built-in--custom)
62	3. [MCP Servers](#mcp-servers-model-context-protocol)
63	4. [Subagent Orchestration](#subagent-orchestration)
64	5. [Session Management](#session-management)
65	6. [Permission Control](#permission-control)
66	7. [Filesystem Settings](#filesystem-settings)
67	8. [Message Types & Streaming](#message-types--streaming)
68	9. [Error Handling](#error-handling)
69	10. [Known Issues](#known-issues-prevention)
70
71	---
72
73	## Core Query API
74
75	### The `query()` Function
76
77	The primary interface for interacting with Claude Code CLI programmatically.
78
79	```typescript
80	import { query } from "@anthropic-ai/claude-agent-sdk";
81
82	const response = query({
83	prompt: string \| AsyncIterable<SDKUserMessage>,
84	options?: Options
85	});
86
87	// Response is AsyncGenerator<SDKMessage, void>
88	for await (const message of response) {
89	// Process streaming messages
90	}
91	```
92
93	### Basic Options
94
95	```typescript
96	const response = query({
97	prompt: "Review this code for bugs",
98	options: {
99	model: "claude-sonnet-4-5", // or "haiku", "opus"
100	workingDirectory: "/path/to/project",
101	systemPrompt: "You are a security-focused code reviewer.",
102	allowedTools: ["Read", "Grep", "Glob"],
103	disallowedTools: ["Write", "Edit", "Bash"],
104	permissionMode: "default" // or "acceptEdits", "bypassPermissions"
105	}
106	});
107	```
108
109	### Model Selection
110
111	\| Model \| ID \| Best For \| Speed \| Capability \|
112	\|-------\|-----\|----------\|-------\|------------\|
113	\| Haiku \| `"haiku"` \| Fast tasks, monitoring \| Fastest \| Basic \|
114	\| Sonnet \| `"sonnet"` or `"claude-sonnet-4-5"` \| Balanced \| Medium \| High \|
115	\| Opus \| `"opus"` \| Complex reasoning \| Slowest \| Highest \|
116	\| Inherit \| `"inherit"` \| Use parent model \| - \| - \|
117
118	Default: `"sonnet"` if not specified
119
120	### System Prompts
121
122	```typescript
123	const response = query({
124	prompt: "Implement user authentication",
125	options: {
126	systemPrompt: `You are an expert backend developer.
127
128	Follow these principles:
129	- Always use TypeScript with strict types
130	- Implement comprehensive error handling
131	- Add detailed logging for debugging
132	- Write unit tests for all functions
133	- Follow OWASP security guidelines`
134	}
135	});
136	```
137
138	CRITICAL:
139	- System prompt sets agent behavior for entire session
140	- Should be clear and specific
141	- Can be 1-10k tokens (affects context window)
142
143	### Working Directory
144
145	```typescript
146	const response = query({
147	prompt: "Refactor the user service",
148	options: {
149	workingDirectory: "/Users/dev/projects/my-app",
150	// Agent operates within this directory
151	// Relative paths resolved from here
152	}
153	});
154	```
155
156	Best Practices:
157	- Use absolute paths for clarity
158	- Agent stays within this directory scope
159	- Critical for multi-project environments
160
161	---
162
163	## Tool Integration (Built-in + Custom)
164
165	### Built-in Tools
166
167	The SDK provides access to Claude Code's

People who liked this also viewed

6 picks

Type

claude-agent-sdk

15stars7forks

View on GitHub

How to install claude-agent-sdk?

jackspace/claudeskillz/claude-agent-sdk

$npx -y skills add jackspace/ClaudeSkillz --skill claude-agent-sdk

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Claude Agent SDK
2
3	Status: Production Ready
4	Last Updated: 2025-10-25
5	Dependencies: @anthropic-ai/claude-agent-sdk, zod
6	Latest Versions: @anthropic-ai/claude-agent-sdk@0.1.0+, zod@3.23.0+
7
8	---
9
10	## Quick Start (5 Minutes)
11
12	### 1. Install SDK
13
14	```bash
15	npm install @anthropic-ai/claude-agent-sdk zod
16	```
17
18	Why these packages:
19	- `@anthropic-ai/claude-agent-sdk` - Main Agent SDK
20	- `zod` - Type-safe schema validation for tools
21
22	### 2. Set API Key
23
24	```bash
25	export ANTHROPIC_API_KEY="sk-ant-..."
26	```
27
28	CRITICAL:
29	- API key required for all agent operations
30	- Never commit API keys to version control
31	- Use environment variables
32
33	### 3. Basic Query
34
35	```typescript
36	import { query } from "@anthropic-ai/claude-agent-sdk";
37
38	const response = query({
39	prompt: "Analyze the codebase and suggest improvements",
40	options: {
41	model: "claude-sonnet-4-5",
42	workingDirectory: process.cwd(),
43	allowedTools: ["Read", "Grep", "Glob"]
44	}
45	});
46
47	for await (const message of response) {
48	if (message.type === 'assistant') {
49	console.log(message.content);
50	}
51	}
52	```
53
54	---
55
56	## The Complete Claude Agent SDK Reference
57
58	## Table of Contents
59
60	1. [Core Query API](#core-query-api)
61	2. [Tool Integration](#tool-integration-built-in--custom)
62	3. [MCP Servers](#mcp-servers-model-context-protocol)
63	4. [Subagent Orchestration](#subagent-orchestration)
64	5. [Session Management](#session-management)
65	6. [Permission Control](#permission-control)
66	7. [Filesystem Settings](#filesystem-settings)
67	8. [Message Types & Streaming](#message-types--streaming)
68	9. [Error Handling](#error-handling)
69	10. [Known Issues](#known-issues-prevention)
70
71	---
72
73	## Core Query API
74
75	### The `query()` Function
76
77	The primary interface for interacting with Claude Code CLI programmatically.
78
79	```typescript
80	import { query } from "@anthropic-ai/claude-agent-sdk";
81
82	const response = query({
83	prompt: string \| AsyncIterable<SDKUserMessage>,
84	options?: Options
85	});
86
87	// Response is AsyncGenerator<SDKMessage, void>
88	for await (const message of response) {
89	// Process streaming messages
90	}
91	```
92
93	### Basic Options
94
95	```typescript
96	const response = query({
97	prompt: "Review this code for bugs",
98	options: {
99	model: "claude-sonnet-4-5", // or "haiku", "opus"
100	workingDirectory: "/path/to/project",
101	systemPrompt: "You are a security-focused code reviewer.",
102	allowedTools: ["Read", "Grep", "Glob"],
103	disallowedTools: ["Write", "Edit", "Bash"],
104	permissionMode: "default" // or "acceptEdits", "bypassPermissions"
105	}
106	});
107	```
108
109	### Model Selection
110
111	\| Model \| ID \| Best For \| Speed \| Capability \|
112	\|-------\|-----\|----------\|-------\|------------\|
113	\| Haiku \| `"haiku"` \| Fast tasks, monitoring \| Fastest \| Basic \|
114	\| Sonnet \| `"sonnet"` or `"claude-sonnet-4-5"` \| Balanced \| Medium \| High \|
115	\| Opus \| `"opus"` \| Complex reasoning \| Slowest \| Highest \|
116	\| Inherit \| `"inherit"` \| Use parent model \| - \| - \|
117
118	Default: `"sonnet"` if not specified
119
120	### System Prompts
121
122	```typescript
123	const response = query({
124	prompt: "Implement user authentication",
125	options: {
126	systemPrompt: `You are an expert backend developer.
127
128	Follow these principles:
129	- Always use TypeScript with strict types
130	- Implement comprehensive error handling
131	- Add detailed logging for debugging
132	- Write unit tests for all functions
133	- Follow OWASP security guidelines`
134	}
135	});
136	```
137
138	CRITICAL:
139	- System prompt sets agent behavior for entire session
140	- Should be clear and specific
141	- Can be 1-10k tokens (affects context window)
142
143	### Working Directory
144
145	```typescript
146	const response = query({
147	prompt: "Refactor the user service",
148	options: {
149	workingDirectory: "/Users/dev/projects/my-app",
150	// Agent operates within this directory
151	// Relative paths resolved from here
152	}
153	});
154	```
155
156	Best Practices:
157	- Use absolute paths for clarity
158	- Agent stays within this directory scope
159	- Critical for multi-project environments
160
161	---
162
163	## Tool Integration (Built-in + Custom)
164
165	### Built-in Tools
166
167	The SDK provides access to Claude Code's

People who liked this also viewed

6 picks

Type