Agent Skills adds subagent creators

The new cursor-subagent-creator skill allows agents to spawn specialized sub-tasks, moving beyond simple chat interfaces. The project's human-curated progressive disclosure model reduces context window bloat by loading instructions only when triggered. Mandatory Snyk scanning and static analysis address the high vulnerability rate found in public AI agent skill ecosystems. High growth on GitHub signals a maturing market for modular, cross-agent capabilities via the SKILL.md standard.