Skills

Agent Setup

Oblien skills follow the AgentSkills open standard. Any agent that reads SKILL.md files can use them. This page covers setup for popular agents.


Claude Code

Claude Code discovers skills from the .claude/ directory in your project root.

Install

npx oblien skills install oblien-runtime --agent claude

Or manually:

mkdir -p .claude/skills/oblien-runtime
curl -sL https://oblien.com/skills/oblien-runtime/SKILL.md \
  -o .claude/skills/oblien-runtime/SKILL.md

Skill locations

LocationScope
.claude/skills/Project-level (commit to git)
~/.claude/skills/User-level (all projects)

Referencing in CLAUDE.md

You can also reference skills in your project's CLAUDE.md to ensure Claude Code loads them:

# Project Instructions

This project uses Oblien workspaces for isolated execution environments.

## Skills

See `.claude/skills/oblien-runtime/SKILL.md` for Runtime API reference.
See `.claude/skills/oblien-networking/SKILL.md` for network configuration.
npx oblien skills install oblien-runtime oblien-networking --agent claude

This gives Claude Code full workspace control - file operations, command execution, and network configuration.


Cursor

Cursor loads context from .cursor/rules/ and supports AgentSkills discovery.

Install

npx oblien skills install oblien-runtime --agent cursor

Or manually:

mkdir -p .cursor/skills/oblien-runtime
curl -sL https://oblien.com/skills/oblien-runtime/SKILL.md \
  -o .cursor/skills/oblien-runtime/SKILL.md

Skill locations

LocationScope
.cursor/skills/Project-level
~/.cursor/skills/User-level (all projects)
npx oblien skills install oblien-runtime oblien-api --agent cursor

OpenClaw

OpenClaw supports AgentSkills natively via clawhub.

Install

clawhub install oblien-runtime

Skill locations

LocationScope
<workspace>/skills/Project-level (local)
~/.openclaw/skills/User-level (global)
Bundled skillsBuilt into OpenClaw

Configuration

Configure which skills are active in ~/.openclaw/openclaw.json:

{
  "skills": {
    "entries": [
      { "name": "oblien-runtime", "enabled": true },
      { "name": "oblien-networking", "enabled": true },
      { "name": "oblien-api", "enabled": true }
    ]
  }
}

Skill gating

Gate skills on environment requirements so they only activate when credentials are present:

metadata:
  openclaw:
    requires:
      env:
        - OBLIEN_CLIENT_ID
        - OBLIEN_CLIENT_SECRET

Hot reload

OpenClaw watches skills directories - update a SKILL.md and it reloads automatically.


Goose

Goose by Block loads skills from .goose/skills/.

Install

npx oblien skills install oblien-runtime --agent goose

Or manually:

mkdir -p .goose/skills/oblien-runtime
curl -sL https://oblien.com/skills/oblien-runtime/SKILL.md \
  -o .goose/skills/oblien-runtime/SKILL.md

JetBrains AI

JetBrains AI Assistant supports AgentSkills from .jb-ai/skills/.

Install

npx oblien skills install oblien-runtime --agent jetbrains

Generic / Custom Agents

For any agent that reads markdown context files, place skills wherever the agent looks for instructions:

# Install to a custom path
npx oblien skills install oblien-runtime --path ./my-agent/context/

# Or download directly
mkdir -p ./my-agent/context/oblien-runtime
curl -sL https://oblien.com/skills/oblien-runtime/SKILL.md \
  -o ./my-agent/context/oblien-runtime/SKILL.md

The SKILL.md is plain markdown with YAML frontmatter. Any agent that can read text files can consume it.


Token overhead

Each installed skill adds to the agent's context:

StageCostWhen
Discovery~100 tokens per skillAt startup (name + description only)
Activation500–5000 tokens per skillOn demand when the skill is relevant

With all 9 Oblien skills installed, discovery overhead is under 1000 tokens. Full skill bodies load only when the agent activates them for a task.

Use caseSkills
Agent working inside a workspaceoblien-runtime
Agent managing multiple workspacesoblien-api + oblien-networking
Agent needing full platform controloblien-runtime + oblien-api + oblien-networking
Agent onboarding to the platformoblien-platform

Most agents only need oblien-runtime + oblien-api. The runtime skill teaches workspace operations (files, exec, search) with platform context, and the API skill teaches workspace management (create, start, stop, scale).