IonBin

Independent reviews. IonBin has no partnerships with the tools on this page and was not paid for these listings. Our evaluations are our own, based on public documentation and our research. Product names are trademarks of their owners and are used for identification only. How disclosure works.

Concept

Agent skills (SKILL.md)

Verified Oct 17, 2026 Review by Apr 15, 2027 3 sources
Short answer

An agent skill is a folder with a SKILL.md file: a short name and description, step-by-step instructions, and optional scripts or reference files. The agent loads it only when a task matches. Agent Skills is an open format that Claude and OpenAI Codex both read, so one written procedure can travel between tools.

What it is

A skill is a written procedure packaged for an AI agent. Think of it as an SOP the agent can open on its own: what the task is, when to use it, the steps, and any templates or scripts it needs.

How it works

  1. A folder per skill. It holds a SKILL.md file with YAML front matter (at minimum a name and a description) followed by plain Markdown instructions. The open specification limits name to 64 characters and description to 1,024 (checked 2026-10-17).
  2. Optional extras. Scripts, reference documents and assets sit beside it.
  3. Loaded on demand. The agent sees the short description first and reads the full instructions only when a task calls for them, which keeps the context window light.
  4. Kept in Git. Store skills in a GitHub repo so changes are reviewed, versioned and shared like code.

Benefits

  • Write once, reuse everywhere. Update the skill and every run uses the new version. That is the property paper SOPs never had.
  • Portable. The open specification is supported by Claude and Codex, so you are not locked into one vendor’s custom-assistant format.
  • A better review question (IonBin’s view). “Which of our repeatable tasks are written down as skills?” is a practical way to audit AI adoption on a team.

Limits and risks

  • Scripts run with the agent’s permissions. Treat third-party skills like any software you install: read them first.
  • Vendor extensions break portability. Stick to the standard front-matter fields in skills you want to share.
  • Instructions drift. A skill nobody owns goes stale. Give each one an owner and a review date.

Checked specs

Snapshot from official sources on Oct 17, 2026.

Required SKILL.md fields (checked 2026-10-17)name (1-64 characters, lowercase letters, numbers and hyphens) and description (1-1024 characters) [1]
LoadingClaude uses progressive disclosure: skill metadata first, full instructions only when relevant [2]
CodexBuilds on the open agent skills standard; the initial skills list is capped at about 2% of the context window or 8,000 characters [3]

Best-fit stacks

Stack 08

Dev, web & IT

Design, code, host and ship sites and internal tools, with an AI editor doing most of the typing.

Agent compatibility

Tool in this entryHow agents connect
Claude Agent platform API/CLIBulk import
OpenAI Codex Agent platform API/CLI
GitHub MCPAPI/CLI

Levels come from each tool's recorded, sourced compatibility. How the levels work.

Sources

  1. Agent Skills specification · Agent Skillsofficial accessed Oct 17, 2026
  2. Agent Skills overview · Anthropicofficial accessed Oct 17, 2026
  3. Skills in Codex · OpenAIofficial accessed Oct 17, 2026

Practitioner views are restated in our words and labeled as views, not measured results. Product facts come from official sources dated above.

Suggest an edit

IonBin's wiki isn't user-edited. Corrections go to our Directory Curator, who checks them against primary sources before anything changes. Spot something wrong or out of date on “Agent skills (SKILL.md)”? Email us with a link to a primary source.

Suggest an edit by email or write to [email protected]