Agent Skills — The Cross-Vendor Format for LLM Agent Task Instructions
Anthropic open-standardized Agent Skills in December 2025; OpenAI Codex and ChatGPT adopted the same format. 17 official skills + 7 commercial partners + enterprise admin controls.
What Skills Are + Their Relation to Plugins and MCP
A Skill is a directory containing SKILL.md (natural-language task instructions) plus optional scripts, templates, and reference assets. The LLM dynamically loads the skill when appropriate to learn how to perform a task class (generate invoices, build slide decks, run a CI report, etc.). Different from Plugins (client-feature packages) and MCP (external tool protocol) — Skills aren't tools or client capabilities; they're 'how to do it' instruction bundles. One plugin can carry Skills and reference MCP servers.
SKILL.md Spec + Minimal Example
SKILL.md uses YAML frontmatter to declare name / description / when-to-use, with the body in markdown for instructions. Example: --- name: invoice-generator description: Generate PDF invoices compliant with China tax regulations using reportlab when-to-use: When the user asks to issue an invoice or generate an invoice --- ## Steps 1. Extract buyer/seller/items from user message 2. Call ./scripts/render.py to output PDF ... The model loads this skill automatically when it sees an invoice-related request.
17 Official Skills + Commercial Partners
The anthropics/skills GitHub repo open-sources 17 Anthropic-maintained skills covering creative design, document creation, technical development, and enterprise communication. Atlassian / Canva / Cloudflare / Figma / Notion / Ramp / Sentry have published commercial skills. The community aggregator SkillsMP (skillsmp.com) collects third-party skills for discovery.
Cross-Vendor Compatibility + Enterprise Admin
After Anthropic open-standardized Skills in Dec 2025, OpenAI Codex CLI and ChatGPT adopted the exact same format. The same SKILL.md runs in Claude Code, Codex CLI, and ChatGPT — write once, deploy across vendors. Claude Team / Enterprise plans expose admin central settings to control which skills users can access and which are enabled by default.
FAQ
Are Skills the same as Custom Subagents?
No. A custom subagent is a Claude Code dispatch unit with its own prompt, tools, and dispatch rules. A Skill is a higher-level instruction package that can be loaded by a main agent or any subagent. A subagent can reference multiple skills.
Can I use a Skill to enforce that the model avoids a behavior?
Not as a hard constraint — Skills are 'recommendations and instructions' that the model can still deviate from under pressure. Hard constraints need hooks (intercept tool calls), permissions (refuse execution), or system-level prompt injection. Use Skills for 'how to do it right'; put hard rules elsewhere.
Can developers in China use Anthropic's official Skills?
Yes. Skills are markdown text repositories — git clone and use locally, no network dependency. Paired with QCode's gateway to Claude/Codex/GPT, you get the same experience as overseas users.
Use Agent Skills Across Vendors via QCode
QCode bundles Claude / GPT / Gemini access — all compatible with the Agent Skills standard. Write once, run everywhere.
Start Your QCode Plan