Skip to main content
A skill is a reusable guide that tells an agent how to do a particular kind of task well: how to build a spreadsheet in the virtual machine, how to write in your brand voice, how to run a monthly report the way your team wants it. Skills stay out of the way until they are needed. The agent sees a short catalog of the skills available to it, and loads the full guide only when a task calls for it, so long procedures don’t crowd its instructions or cost tokens on every turn. Skills use the open Agent Skills format, the same one used by Claude and a growing set of other tools. A skill is a folder with a SKILL.md file that starts with a short header giving the skill’s name and when to use it, followed by the instructions themselves. It can also bundle supporting files such as reference documents, templates and scripts, which the agent reads on demand. In Abundly, the skills you add to an agent are ordinary documents in a folder called skills, so they get everything documents have: version history, the document editor, protection, and a place on the agent’s Docs page.

​
The Skills tab

Open Capabilities → Skills on an agent to see everything it can load, in two groups:
  • Added to this agent are the skills you or a colleague added to this agent. Skills belong to the agent: another agent in the workspace has its own list.
  • Built-in skills come with the platform. Some are general; others belong to a capability, such as the virtual machine or Slack, and appear here once that capability is enabled. The card is tagged with the capability it came with.
Click a built-in skill to read it. Click one of the agent’s own skills to open it in the document editor, where you can change its text, its description and its name. Use the switch on a card to turn a skill off for this agent: the agent will neither see it in its catalog nor be able to load it. This is useful when a built-in skill pulls the agent in a direction you don’t want for a particular job.
Turning off a skill only affects this agent. Other agents in the workspace keep their own settings.

​
Adding skills to an agent

You need Edit access to the agent to add or remove skills. Add from URL takes a link to the skill. Any of these work:
  • A folder on GitHub that contains a SKILL.md, for example a skill in the Anthropic skills repository or in your own repository. Supporting files in the folder are included.
  • A direct link to a SKILL.md file.
  • A link to a .skill or .zip archive of the skill folder.
Upload takes a .skill or .zip archive, or a single SKILL.md, from your computer. Skills exported from Claude use the .skill format and can be uploaded as they are. Write opens the document editor on a new file in the skills folder, so you can paste a skill you have or write a short one on the spot. For imports, the skill’s name and description come from the SKILL.md header. If the header is incomplete, the platform derives them from the file and tells you what it used. The header is only used at import: afterwards the document’s name and description are what count. Binary files in a skill, such as images, are left out; text files such as reference notes and scripts come along.

​
Skills are documents

Every skill added to an agent lives in a folder called skills at the top of the agent’s documents. Anything placed directly in that folder is a skill:
  • A sub-folder that contains a SKILL.md, plus any supporting files. This is what Add from URL and Upload create when the skill brings files along.
  • A single file, usually markdown, that is the guide itself. This is what Write creates, and what an import creates for a skill without supporting files.
Name the document whatever you like; the agent loads it by a short id derived from the name. The document’s description is what the agent reads to decide whether the skill applies, so keep it specific about both what the skill does and when to use it. Because these are documents, you can also create a skill from the Docs page by adding a file or folder to the skills folder, move one between agents, or protect it like instructions. A sub-folder without a SKILL.md is not a skill yet. The on/off switch on the Skills tab hides or shows the skill to the agent, the same as the document’s agent visibility setting. Hiding the skills folder itself turns off all of them. A skill of the agent’s own with the same id as a built-in skill replaces it for that agent: the built-in card shows a “Replaced” badge, and the agent loads yours instead. This is how you tailor a built-in guide to your own way of working. Once added, the skill is on and listed in the agent’s catalog right away. There is no need to change the agent’s instructions, though it often helps to mention in the instructions when the agent should reach for a skill.
A skill is a set of instructions the agent will follow. Read a skill from a third party before you add it, just as you would review a document you hand to a new colleague, and only take skills from sources you trust.

​
How the agent uses skills

In every conversation the agent sees a one-line summary of each available skill. When a task matches one, it loads the full guide, and any supporting file it needs, and follows it. Agents with the Shell capability also see every skill as a file under ~/documents/skills/, built-in ones and their own. If you want the agent to always follow a guide rather than load it on demand, put the guide in its instructions or in a protected document instead. Skills are for procedures that only matter some of the time.

​
Writing a skill

A minimal SKILL.md:
The description matters most: it is what the agent reads to decide whether the skill applies, so say both what the skill does and when to use it. Keep SKILL.md focused on the procedure and move long reference material into separate files in the folder; the agent loads those only when the guide points to them. When you write a skill directly in Abundly you can skip the header and fill in the document’s description instead.