Writing SKILL.md files for Claude Code, Cursor, and other agents

On 2026-10-23, an article on Juejin described integrating Cloudflare's open source security-audit-skill into Claude Code. It framed the move around Anthropic's donation of Agent Skills to the AAIF and walked through the six-stage workflow of Cloudflare's project, which has around 37k stars on GitHub. Rather than scanning everything, the author recommended trimming the skill's scope to only what a task needs, which keeps token usage down.

Early on 2026-10-24, a post on DEV Community's AI channel, titled "How to Write SKILL.md for Claude Code, Cursor, and More," took a broader view. It argued that Agent Skills have become the standard way to teach an AI coding agent a repeatable job: write a single SKILL.md folder and Claude Code, Cursor, OpenAI Codex, GitHub Copilot, and VS Code can all load it. It added that Next.js 16.4 now ships its own skills, installable with npx skills add. To demonstrate, the author built next-security-check, a skill that checks a Next.js app against the October 2026 security advisories. The skill was validated with the official skills-ref validator and installed via the skills CLI. Along the way the guide covers formatting rules, common triggering mistakes, and references to the official documentation.

Taken together, the two reports show the same theme from two angles: one is a concrete onboarding recipe for an existing third-party skill, the other a how-to for authoring your own. The materials contain no later developments, so the record ends at this pair of posts.