skillevaluator
NVIDIA open-sourced a framework that automatically tests whether an AI agent's skills actually make it behave better, not just whether the skill files look correct.
View nvidia/skillevaluator on GitHub- GitHub stars
- 1,468
- Forks
- —
- Language
- Python
Project details
- Repository
- nvidia/skillevaluator
- Licence
- Apache-2.0
- Primary language
- Python
- Catalog status
- Available
- Catalog check
- 2026-11-15
Repository description
Multi-tier framework for evaluating AI agent skills with quality gates, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation that measures how skills affect agent behavior.
Explore by topic
More tools in Skills & plugins
- skillspector
nvidia · Apache-2.0
Scans AI agent skills for Claude Code, Codex, and MCP for vulnerabilities and malicious patterns before you install them, after finding that over a quarter of a large…
- agent-skills
addyosmani · MIT
A packaged set of production-grade engineering skills that walk AI coding agents through the senior-engineer workflow: define, plan, build, test, review, ship.
- paperclip
paperclipai · MIT
An open-source dashboard where you hire a team of AI agents (Claude Code, Codex, Cursor, OpenClaw, etc.), give them a business goal, set budgets, and watch them work…
- agents
wshobson · MIT
A one-stop marketplace of ready-made AI coding agents, skills, and commands you can install into Claude Code, Cursor, Codex, or five other AI coding tools with a single…
