SkillOS — Skill Operating System (v0.6.0-beta)

Agent skill scheduling and orchestration kernel.

SkillOS is the terminal implementation layer of the OpenHuman Protocol — a Python 3.11+ backend for multi-agent skill execution with a built-in 5/25/70 economic model.

Quick Start

uv sync --extra dev
uv run pytest

Architecture

L3 经济层    5/25/70 分配引擎 · Skill Market L3 MVP · NegMarket 负样本市场
             生态基金 · Registry / Pricing / Settlement
L2 协议层    Registry / Matcher / Settlement / Identity / Remote Bridge
L1 内核层    EntryKernel · AgentScheduler · SkillKernel · SessionManager
             SessionIsolator · StatePersistor · TaskScheduler · WebFetcher
L0 基础层    结构化日志 · Prometheus 指标 · OTel 追踪 · 错误码系统

See ARCHITECTURE.md for details.

Development

make test     # run all tests
make lint     # ruff check
make audit    # full release gate

Inspiration

The intelligence layer (skillos.intelligence) design was inspired by concurrent academic research on skill self-evolution (independently implemented, no code or data shared):

• SkillEvolver: Skill Learning as a Meta-Skill — Genrui Zhang et al., arXiv:2605.10500, May 2026
• EmbodiSkill: Skill-Aware Reflection for Self-Evolving Embodied Agents — Ruofei Ju et al., arXiv:2605.10332, May 2026

SkillOS's evolution loop uses its own Agent execution logs and architecture vocabulary, not derived from any external codebase.

License

AGPL-3.0 — see LICENSE.