skill.md

Agent skills for EigenCloud. Drop these into any AI coding agent and give it the ability to deploy, manage, and operate verifiable applications on EigenCompute TEEs.

What's in here

Skill	What it does
ecloud	Full lifecycle management for EigenCompute TEEs — auth, billing, deploy, verifiable builds, operate, upgrade, rollback, terminate. Any agent that loads this skill can deploy to a TEE, manage USDC credits, verify provenance, and handle the full app lifecycle.

What an agent can do with this

Once the ecloud skill is loaded, an agent can:

• Deploy to TEEs — pre-built images or verifiable builds from git source
• Handle billing — subscribe, top up USDC credits, check balance
• Manage running TEEs — start, stop, upgrade, view logs, check wallet addresses
• Verify provenance — verifiable builds with onchain build hash recording
• Rollback — find previous releases and upgrade back to a prior image
• Terminate TEEs safely — with fund withdrawal warnings and confirmation flows
• Troubleshoot — Docker platform targeting, port binding, auth failures, name vs hex ID issues

The skill defaults to keeping everything private — log visibility, environment variables, and app configuration are all private unless explicitly opted into public.

Install

Install this skill: https://github.com/Layr-Labs/skill.md

Works with Claude Code, Codex, OpenClaw, and any agent that supports the Agent Skills open standard.

Skill structure

ecloud/
├── SKILL.md                        # Main skill — gate-based runbook for deployment lifecycle
└── references/
    ├── cli-reference.md            # Complete ecloud CLI command reference
    └── architecture.md             # TEE architecture, KMS, security model

Prerequisites

The skill expects the user's machine to have:

• Docker installed and logged in (docker login)
• ecloud CLI installed (npm install -g @layr-labs/ecloud-cli)
• ETH for deployment gas (Sepolia testnet or mainnet)

The skill walks the agent through checking and setting these up if they're missing.

Contributing

Add new skills as top-level folders with a SKILL.md. Keep the same structure:

my-skill/
├── SKILL.md              # Required — name, description, workflows
└── references/           # Optional — supporting docs

Frontmatter format:

---
name: "skill-name"
description: "One-liner for when the agent should activate this skill."
---

The description field matters — agents use it to decide whether to load the skill for a given task. Be specific about triggers and scope.