What is SPM?

The open, developer-first registry for battle-tested procedural skill packages designed specifically for autonomous AI agents.

Compounding Agent Capabilities

AI agents are powerful, but developers spend too much time building the exact same tool definitions, sandbox runtimes, and validation schemas from scratch.

SPM (Skill Package Manager) is the dependency manager for agent autonomy. It acts as the npm or pip of AI skills, enabling you to discover, import, version, and orchestrate complex procedural tasks (like secure Postgres querying, headless browser automation, or calendar booking) in seconds.

How it works

1Discover & Pull

Find battle-tested skill schemas on the registry and fetch them using the SPM CLI.

2Import & Bind

Load package code directly inside LangChain, LlamaIndex, or custom agent systems.

3Execute Safely

Every skill provides isolated runtimes, strict input validation, and clear outputs.

A Quick Demonstration

Install and initialize packages using a single shell routine:

bash

# Install a package from the SPM registry
spm install @utils/json-validator

# Initialize a package in your local codebase
spm init @utils/json-validator --namespace custom-skills

Core Pillars of SPM

Sandboxed Autonomy

Every downloaded package complies with standard execution layers. Run dangerous browser actions or data parser queries within safely managed environments.

Semantic Schemas

Skills provide strict OpenAPI or custom JSON schemas so your LLM agents understand precisely what tools exist, their requirements, and expected payloads.

Framework Agnostic

Integrate downloaded skills intoLangChain, CrewAI, AutoGen, LlamaIndex, or build your own custom runtime wrappers in pure Python or TypeScript.

Fast & Scalable

Resilient version locking ensures your remote production pipelines never break when packages receive upgrades. Focus purely on model engineering.

NEXT DOCUMENTInstallation & CLI