Metaskills 1.0

Metaskills 1.0: Portable Agent & Skill Building

Metaskills 1.0 brings a unified, portable approach to agent and skill building. This release empowers teams to construct powerful, self-contained AI agents and reusable skills directly within their own projects, eliminating the need for complex centralized platforms or proprietary runtimes.

Intention

Our goal with 1.0 is to decentralize AI agent development. We believe that agent intelligence should live alongside your code, not in an external walled garden. This release formalizes our foundation, establishes robust feedback and evaluation mechanisms, and delivers a seamless developer experience for discovering and integrating skills.

Key Capabilities

🛠️ Agent Builder Skill

Transform plain-language goals into fully configured, portable agents. Whether you need a code reviewer, an onboarding assistant, or a custom tool-enabled copilot, the agent-builder skill automatically generates the specifications, instructions, boundaries, and starter prompts you need to get up and running quickly.

🧪 Skill Creator Skill

Turn your ad-hoc workflows, prompts, and scripts into production-ready, reusable skills. The skill-creator goes beyond generation—it includes built-in capabilities to evaluate, benchmark, and iteratively improve your skills, complete with optional learning and telemetry guidance to ensure long-term stability and performance.

🚀 Developer Experience (DX) & Integrations

• Effortless Installation: Use the standard npx skills add CLI to seamlessly install self-contained skills into your target agent environment (.agents, .claude, etc.).
• Clean Submodule Management: External reference components are now cleanly managed via Git submodules, avoiding code duplication while maintaining tight integration.
• Comprehensive Documentation: Added full documentation for repository architecture and component workflows to streamline adoption and maintenance.

Foundation & Under the Hood

This release marks the completion of three major foundational sprints:

• Sprint 1 - Formalization Foundation: Establishing AGENTS.md as the unified, portable source of truth.
• Sprint 2 - Feedback & Eval Engine: Introducing mechanisms to benchmark and improve existing skills.
• Sprint 3 - Integrations & DX: Polishing the adopter experience with plural skills installer compatibility and refocused onboarding guides.