Skip to content

haiku.skills

Skill-powered AI agents implementing the Agent Skills specification with pydantic-ai.

How it works

SkillToolset is a pydantic-ai FunctionToolset that you attach to your own agent. It supports two execution modes:

  • Sub-agent mode (default) — Exposes a single execute_skill tool. When the agent calls it, a focused sub-agent spins up with only that skill's instructions and tools, then returns the result. Each skill runs in isolation with its own system prompt, tools, and token budget.
  • Direct mode (use_subagents=False) — Exposes query_skill, execute_skill_tool, run_skill_script, and read_skill_resource tools. The main agent calls skill tools directly — no sub-agent LLM loops, lower latency, and the agent retains tool results in its conversation context.

Features

  • Two execution modes — Sub-agent delegation for isolation, or direct tool access for speed and context retention
  • Filesystem skills — Load SKILL.md directories with scripts in Python, JavaScript, TypeScript, or shell
  • Entrypoint skills — Install skill packages with typed in-process tools, per-skill state, and zero-config discovery
  • Per-skill state — Pydantic state models tracked per namespace; state changes emit StateDeltaEvent (JSON Patch) for the AG-UI protocol
  • MCP integration — Wrap any MCP server (stdio, SSE, streamable HTTP) as a skill
  • Signing and verification — Identity-based skill signing via sigstore

Quick install

uv add haiku.skills

Documentation

  • Installation — Install with optional extras
  • Tutorial — From filesystem skills to entrypoint packages, step by step
  • Skills reference — SKILL.md format, tools, state, and script resolution
  • Signing — Identity-based signing and verification
  • CLI — Command-line interface reference
  • AG-UI protocol — State deltas and the AG-UI protocol
  • Example skills — Built-in skill packages as reference implementations
  • Development — Contributing, testing, and tooling