Skills

A Skill packages reusable instructions and supporting files for Agent work. Use a Skill for guidance that should apply repeatedly, such as review rules, delivery checklists, writing standards, or a team process.

Keep one-time requirements in the Task description or comments. Create a Skill when the same capability should be available across future runs.

Skill scopes

Every Skill belongs to an organization and has one scope:

  • Organization: available to eligible Agents throughout the organization.
  • Space: available to Agents associated with that Space. An Agent cannot bind a Space Skill unless the Agent belongs to that Space.
  • Agent: belongs to one Agent and is bound to that Agent when created or imported.

Choose the narrowest scope that matches the intended audience. A Skill's scope cannot be changed after creation; create a new Skill when its audience needs to change.

Create or import a Skill

In Skill or Agent settings, select the unified + Add Skills entry and then choose a source. The available paths include:

  • Select from Space Skills: choose a Skill in the Agent's current Space that the current person is allowed to use; this binds the existing Skill and does not create a copy;
  • Create manually: start with a name, description, and instructions;
  • Upload files or a folder: upload a SKILL.md, a .skill file, a zip package, or a folder containing the Skill files;
  • Import from a URL: import a Skill from a supported public source;
  • Import from a Runtime: choose from Skills discovered by the selected Computer and Runtime.

Skills imported from a URL, Runtime, file, or folder become managed copies in Sharkly. They remain separate from the original URL, local files, or local directory. Skills already available directly to an Agent's Runtime do not need to be bound in the product; import one when it should become a managed Skill with an explicit scope. Review imported content before binding it to an Agent.

A valid Skill normally uses SKILL.md as its entry point, with YAML name and description fields followed by instructions. A zip package or folder can also include templates, examples, and other supporting text files. Keep paths relative to the Skill and do not include passwords, access tokens, private keys, or other secrets. Uploads and imports reject hidden or sensitive paths, unsupported binary content, unsafe archive entries, duplicate paths, and packages that exceed service limits.

Use the Agent's Skills settings to add or remove eligible Skills. You can also select Skills while creating an Agent.

Binding follows scope rules:

  • an Organization Skill can be bound to an Agent in the same organization;
  • a Space Skill can be bound only when the Agent belongs to that Space;
  • an Agent Skill is already associated with its target Agent.

Changes apply to later Agent runs. Avoid binding many overlapping Skills: conflicting or redundant instructions make execution less predictable.

Permissions and maintenance

A Skill can be edited or deleted by its creator or an Organization Owner or Admin. A Space-scoped Skill also requires write access to its Space. Creating or importing an Agent-scoped Skill, and changing an Agent's Skill bindings, requires permission to manage the target Agent.

When a process changes, update the Skill's instructions and supporting files. Deleting a Skill removes it from Agents that use it, so check current bindings before deletion.

Security guidance

  • Never store passwords, access tokens, private keys, or other secrets in Skill instructions or supporting files.
  • Review every external or locally imported Skill before use.
  • Keep commands and examples least-privileged and specific to the intended task.
  • Do not include sensitive organization data unless every user allowed by the Skill's scope may access it.
  • Recheck bindings after changing an Agent's Space associations or a Skill's contents.