Resources
Curated with explicit editorial criteria: primary sources preferred, annotations required, last-verified dates recorded.
documentation
- HazelJS documentation
Canonical docs home for Agent OS, packages, guides, and starters. Start here for every Craft HazelJS path.
HazelJS · foundation · verified 2026-08-13
- HazelJS Agent OS guide
Layered model of Agent OS—kernel, skills, durability, HITL, and operator surfaces. Primary narrative companion to Craft patterns.
HazelJS · intermediate · verified 2026-08-13
- Skillgate documentation
Governed skills from OpenAPI and related sources into ToolRegistry—how HazelJS packages capability kits for agents.
HazelJS · intermediate · verified 2026-08-13
- HazelJS Agent package
AgentRuntime, @Agent/@Tool, durable runs, budgets, PolicyEngine, and loop options—the core API surface for Craft patterns.
HazelJS · intermediate · verified 2026-08-13
- HazelJS MCP package
First-party MCP integration for discovering and mediating tool servers inside HazelJS apps.
HazelJS · intermediate · verified 2026-08-13
- Function calling (OpenAI platform docs)
Provider-level tool/function-calling mechanics useful when wiring model adapters behind HazelJS ToolExecutor.
OpenAI · foundation · verified 2026-08-13
- Tool use (Anthropic docs)
Provider documentation for tool use patterns and message formats—supporting reference for multi-provider HazelJS stacks.
Anthropic · foundation · verified 2026-08-13
observability
- HazelJS Observability package
Tracing and metrics for model/tool spans; pairs with Inspector timelines for AgentRun debugging.
HazelJS · intermediate · verified 2026-08-13
- OpenTelemetry documentation
Vendor-neutral observability standard underpinning @hazeljs/observability spans for model and tool calls.
CNCF OpenTelemetry · intermediate · verified 2026-08-13
security
- HazelJS Guardrails package
Runtime controls for inputs, outputs, and unsafe agent actions—complement PolicyEngine and HITL gates.
HazelJS · intermediate · verified 2026-08-13
- OWASP Top 10 for Large Language Model Applications
Threat categories for LLM apps—prompt injection, supply chain, excessive agency. Use in design reviews for tool-using HazelJS agents.
OWASP · intermediate · verified 2026-08-13
specifications
- Model Context Protocol
Primary specification site for MCP—tool/resource discovery and host-mediated invocation. Supporting interop spec for @hazeljs/mcp.
Model Context Protocol · intermediate · verified 2026-08-13
- Model Context Protocol organization (GitHub)
Official MCP organization for SDKs, servers, and protocol repositories—supporting reference beside @hazeljs/mcp.
GitHub · intermediate · verified 2026-08-13
- JSON Schema specification
Standard for tool argument and structured-output contracts used by @Tool schemas and AgentRuntime validation.
JSON Schema · foundation · verified 2026-08-13
research
- Building effective agents
Practical guidance on when to use workflows vs agents and how to keep architectures simple. Conceptual companion to HazelJS patterns.
Anthropic · foundation · verified 2026-08-13
- 12-Factor Agents
Opinionated engineering factors for agentic applications—complementary checklist mindset alongside Agent OS production guidance.
HumanLayer · intermediate · verified 2026-08-13