Skills

Skills give your agent reusable instructions for a kind of task. Each is a folder following the Agent Skills standard, with a SKILL.md describing its name, purpose and instructions. Add the folder once, then edit its files as your needs change.

Add a skill

Copy a skill from a local folder or a repository URL into your agent's skills collection. The source folder must contain SKILL.md; adding from a URL also needs git.

lnk agent skill add ~/Downloads/pdf
lnk agent skill add https://github.com/example/skills/tree/main/skills/pdf

Where it is saved

The command copies the folder into ~/Link/Agents/<agent>/Agent/skills/<name>/. Edit the copied SKILL.md and its supporting files there. The collection moves with your agent; Configuration describes the folder, and Switch Harnesses explains what another harness receives.

How the harness reads it

A harness decides how skills enter its prompt and how their full instructions are read. Link Harness lists metadata from the next message and reads the selected skill through memory. An adapter may translate skills into its harness's own files, so consult that harness's documentation for its behavior.

See installed skills

List the agent's skills by name and description to verify what was installed. The agent's skills folder is also open to you as ordinary files, so you can review its complete instructions before using it.

lnk agent skill list

Remove a skill

Remove a skill by name when the agent should no longer use its instructions. You can also delete its folder from the skills collection; the original folder you added from is a separate copy.

lnk agent skill remove pdf

Scripts and permissions

A skill may include scripts or instructions to run programs. Installing it grants no extra permission: its scripts run only when the agent executes them through a tool, within the sandbox and program permissions you allowed. Permissions explains how to control that access.