agent skills
7 mentions across all digests
Agent skills are a standardized format for packaging reusable instructions that enable AI agents to follow team processes, quality standards, and domain expertise automatically, as defined by Vercel and supported by tools like Claude Code, Cursor, and Gemma 4 on-device.
CC-Canary: Detect early signs of regressions in Claude Code
Delta HQ releases CC-Canary, a drift detection tool for Claude Code that analyzes session logs locally to surface model regressions before they impact workflows.
Bilevel Optimization of Agent Skills via Monte Carlo Tree Search
Bilevel optimization combined with Monte Carlo Tree Search enables more efficient hierarchical learning of AI agent capabilities through adaptive sampling.
Show HN: Marimo pair – Reactive Python notebooks as environments for agents
Marimo pair integrates reactive Python notebooks as Agent Skills execution environments, giving AI agents direct code execution and system utility access.
Agent skills explained: An FAQ
Vercel standardizes instruction packaging for AI agents with 'agent skills'—a versioned format that eliminates instruction sprawl and enables teams to automatically operationalize processes, standards, and expertise.
Introducing: React Best Practices
Vercel packages 40+ React performance rules as AI-agent-optimized Agent Skills, enabling Claude Code and Cursor to automatically refactor performance patterns at scale.