claude-agent-sdk

SKILL.md

Build autonomous AI agents with structured JSON validation, event-driven hooks, and 14 documented error preventions.

Structured outputs (v0.1.45+) guarantee JSON schema validation with full TypeScript inference via Zod or raw schemas

12 hook events (PreToolUse, PostToolUse, SubagentStart, SubagentStop, etc.) enable event-driven workflows, custom permission logic, and cleanup patterns

MCP server integration supports stdio, HTTP, and SSE transports with tool naming conventions and strict configuration validation

Session management includes forking for safe exploration, V2 preview APIs for simpler multi-turn patterns, and file checkpointing for rollback capability

Permission control via canUseTool callbacks and sandbox settings for secure execution with auto-approval logic and excluded command lists

Prevents 14 documented issues including CLI not found, context length session-breaking, MCP config gotchas (missing type field), Unicode line separator parsing, and subagent cleanup orphaning

Claude Agent SDK - Structured Outputs & Error Prevention Guide

Package: @anthropic-ai/claude-agent-sdk@0.2.12
Breaking Changes: v0.1.45 - Structured outputs (Nov 2025), v0.1.0 - No default system prompt, settingSources required

What's New in v0.1.45+ (Nov 2025)

Major Features:

1. Structured Outputs (v0.1.45, Nov 14, 2025)

JSON schema validation - Guarantees responses match exact schemas

outputFormat parameter - Define output structure with JSON schema or Zod

Access validated results - Via message.structured_output

Beta header required: structured-outputs-2025-11-13

Type safety - Full TypeScript inference with Zod schemas