<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> SkillsVote addresses the challenge of managing a vast and evolving ecosystem of agent skills by providing a lifecycle governance framework. It targets engineers, researchers, and power users working with coding, research, or workflow agents, offering a benefit of improved agent performance on complex, long-horizon tasks through intelligent skill selection and adaptive library evolution.

How It Works

The project treats agent skills as lifecycle-managed artifacts, integrating collection, profiling, just-in-time recommendation, trajectory-based attribution, and feedback-driven evolution into a continuous loop. This approach enables efficient management of over 1.68 million discovered skills, moving beyond static, manually curated lists. Its novelty lies in leveraging grounded, attributed feedback from task execution to evolve reusable skill libraries, enhancing agent capabilities at scale.

Quick Start & Requirements

Two primary integrations are offered: skills-vote for cloud-based recommendation and attribution via the hosted SkillsVote service, and skills-vote-local for local or private skill library recommendation without relying on the hosted index.

• Installation: Manual CLI installation uses npx skills add MemTensor/skills-vote --skill skills-vote (hosted) or npx skills add MemTensor/skills-vote --skill skills-vote-local (local). Agent integration prompts are also provided.
• Prerequisites: The hosted service requires SKILLS_VOTE_API_KEY and GH_TOKEN. The local version requires local configuration (configs/config.yaml).
• Documentation: Installation details are available at https://raw.githubusercontent.com/MemTensor/skills-vote/main/integration/skills/INSTALL.md. Technical reports are on arXiv:2605.18401.

Highlighted Details

• Hosts the world's largest open agent skill library, with over 1.68 million SKILL.md files discovered and 790,000+ format-valid skills.
• Evaluated on agentic coding and terminal challenge benchmarks including Terminal-Bench Pro, Terminal-Bench 2.0, and SWE-Bench Pro, demonstrating performance improvements on long-horizon tasks.
• Recent developments include the release of a technical report (arXiv:2605.18401) and the skills-vote-local integration for private skill recommendation.

Maintenance & Community

The project is developed by MemTensor. Specific community channels (e.g., Discord, Slack) or notable external contributors are not detailed in the README.

Licensing & Compatibility

The repository is licensed under the permissive MIT License, generally allowing for commercial use and integration into closed-source projects.

Limitations & Caveats

The roadmap indicates that local attribution and evolution capabilities, along with the release of main experiment trajectories and results for full reproduction, are still under development and not yet fully integrated or released.