SKILL.md Generator
Generate SKILL.md files for any AI agent — Claude Code, Cursor, Codex, Gemini CLI, and more. Role-based templates, workflow skills, and the full spec.
Unique identifier (lowercase, alphanumeric, dashes)
Agents use this to decide when the skill is relevant
3 Frontmatter Options grouped by portability
Environment requirements (system packages, network access, intended product)
disable-model-invocation: true — skill only activates via explicit /skill-name, never auto-appliedGlob patterns that scope the skill to matching files
Pre-approved tools the agent can use without asking. Support varies by tool.
Extra discovery context appended to the description
Shown during / autocomplete to indicate expected arguments
/ menuuser-invocable: false — background knowledge only the agent loads, not a user command4 Skill Templates Choose a starting point
5 Instructions Customize behavior
6 Custom Instructions Optional
Preview
Where to place your skill
.agents/
└── skills/
└── my-skill/
├── SKILL.md
├── scripts/ ← optional executables
├── references/ ← optional docs
└── assets/ ← optional templatesAuto-discovered from:
.agents/skills/Project — cross-platform
.cursor/skills/Project — Cursor
~/.cursor/skills/Global — Cursor
.claude/skills/Project — Claude Code
~/.claude/skills/Global — Claude Code
Frequently asked questions
Quick answers about the SKILL.md format, which tools support it, and how the fields differ across agents.
What is a SKILL.md file?
A SKILL.md is the entry point of an Agent Skill — a folder that packages reusable instructions, and optional scripts, references, and assets, for an AI coding agent. It has YAML frontmatter (at minimum a name and description) followed by a Markdown body the agent loads only when the skill is relevant. The format was created by Anthropic and published as an open standard in December 2025.
Does SKILL.md work outside Claude?
Yes. SKILL.md is a cross-vendor open standard. Claude Code, Cursor, OpenAI Codex, and Gemini CLI all read SKILL.md folders. Cursor even loads .claude/skills/ and .codex/skills/ for compatibility, and Gemini CLI and Codex both honor the shared .agents/skills/ directory. The directory format and the name and description fields are portable everywhere.
Are the frontmatter fields the same across every tool?
No — only the core fields are universal. name, description, metadata, license, and compatibility work everywhere. disable-model-invocation and paths are supported by Claude and Cursor but not by Codex or Gemini frontmatter (Codex uses an agents/openai.yaml sidecar instead). Fields like when_to_use, argument-hint, user-invocable, and context: fork are Claude Code extensions. This generator groups fields by portability so you know which are safe to rely on.
How is a skill different from CLAUDE.md or AGENTS.md?
CLAUDE.md and AGENTS.md hold always-on project facts that load every session. A skill is on-demand: its body loads only when the task matches its description, so long procedures and reference material cost almost no context until they are needed. Use AGENTS.md or CLAUDE.md for persistent project guidance, and a skill for a repeatable workflow or specialized capability.
Where do skill folders live?
In a skills directory that the agent auto-discovers. The cross-platform location is .agents/skills/ (project) or ~/.agents/skills/ (global). Tool-specific paths include .cursor/skills/ and ~/.cursor/skills/ for Cursor, .claude/skills/ and ~/.claude/skills/ for Claude Code, and .gemini/skills/ for Gemini CLI. Each skill is its own folder whose name must match the name field in the frontmatter.
Is anything I type sent to a server?
No. The generator runs entirely in your browser. Your skill name, description, and frontmatter never leave the page, and the tool works offline once loaded.