Skills

Overview

Skills are reusable instructions, distinct from executable tools. Invoke one with /skill-name in chat; Arden loads its SKILL.md before working on the request.

Builtin skills

Skill	Use
`/add-model`	Register an OpenAI-compatible chat or embedding model
`/add-skill`	Create or remove a project or user skill
`/add-tool`	Scaffold a user-defined Python tool
`/audit`, `/investigate`, `/panel`, `/implement`	Run a curated multi-agent workflow through `workflow`
`/loop`	Schedule repeated work in the current chat
`/mermaid`	Return a renderable Mermaid diagram
`/propose-automation`, `/propose-skill`	Propose a reusable automation or skill from this chat
`/wiki-automation`	Create an automation that writes managed wiki output

The four workflow skills are trusted built-ins only; user-authored Python workflow presets do not run.

Skill locations

Skills are discovered in this order; the first matching name wins:

apps/server/skills/ – built-ins when developing from source; packaged as skills/ in installed builds
agent/skills/ – filesystem-agent skills
.agents/skills/ – project skills shared with agent clients
.skills/ – Arden project-level skills (per-directory)
~/.arden/skills/ – global user skills
~/.agents/skills/ – global skills shared with other agent clients

Installing skills

Install a skill from GitHub:

curl -X POST http://localhost:6877/skills/install \
  -H "Authorization: Bearer $ARDEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"source": "owner/repo/path/to/skill"}'

Or ask the agent:

Install the skill from github.com/user/repo/my-skill

Creating a skill

A skill is a directory containing at minimum SKILL.md:

my-skill/
├── SKILL.md       # Description, parameters, instructions
└── run.py         # Optional executable

The SKILL.md tells the agent what the skill does and how to execute it. The agent follows these instructions when the user invokes /my-skill. Restart the server after creating a skill manually. Skills installed through the API reload automatically.

Governance

Skill metadata is validated at registry load time. name must match the directory name and use lowercase letters, digits, and hyphens. description is required. Optional governance fields:

Field	Meaning
`source`	Where the skill came from, such as `github:owner/repo/path` or `local`
`version`	Version, date, commit, or release label reviewed by the user.
`reviewed_at`	ISO date for the last human review.

GET /skills/governance returns inventory, validation issues, and cleanup candidates. Cleanup candidates are recommendations only; Arden removes a skill only through the explicit delete action.

Managing skills

Command	Description
`GET /skills`	List installed skills
`DELETE /skills/{name}`	Uninstall a skill

Getting Started

Features

Integrations

Reference

Overview

Builtin skills

Skill locations

Installing skills

Creating a skill

Governance

Managing skills

​Overview

​Builtin skills

​Skill locations

​Installing skills

​Creating a skill

​Governance

​Managing skills

Overview

Builtin skills

Skill locations

Installing skills

Creating a skill

Governance

Managing skills