daocloud-skills

A generated CLI and AI skill package for DaoCloud Enterprise (DCE). It wraps the DCE REST API into a dce command-line tool and bundles a companion skill for AI agents.

Overview

• dce — a CLI generated from DCE OpenAPI specs across supported modules. Supports API discovery, search, and execution with built-in auth management.
• skills/dce — an AI agent skill that teaches agents how to use dce safely.

Currently supported products:

Prerequisites

• Go 1.25+
• Git (for spec sync)
• Docker with buildx (for container image builds)

Quick Start

Quick Start is for contributors building from source. If you only want to use the CLI, skip to Install the CLI.

# Sync OpenAPI specs and regenerate code
make bootstrap

# Build dce
make build

# Install dce and symlink skill for local development
make dev

The make dev target installs dce to /usr/local/bin and symlinks skills/dce into ~/.agents/skills/dce. After make dev, you can jump straight to Usage; you do not need to run “Install the Skill” or “Install the CLI” below.

Install the Skills

Only needed if you want to install this repository’s skills into an AI agent environment without running make dev. This installs the skills only (it does not install the dce CLI binary).

Install all skills globally (--skill '*' selects every skill in the repo):

npx skills add daocloud/daocloud-skills --skill '*' -g

For non-interactive installs, add confirmation flags:

npx -y skills add daocloud/daocloud-skills --skill '*' -g -y

The -g flag installs the skills into the selected agent’s user-level skills directory, so they are available across projects. Omit -g if you want to install them only for the current project.

This installs the skills only. They help an AI agent operate DCE, but do not install the dce CLI. Install the CLI separately before running DCE operations.

--skill '*' installs all skills in the repo. To install only specific ones, pass their names instead, e.g. --skill dce (run the --list command below to see the available names).

Install for a Specific Agent

The CLI auto-detects installed agents and installs to them. Use --agent to target specific AI coding agents instead. Common client targets include:

Claude Code

npx skills add daocloud/daocloud-skills --skill '*' -g --agent claude-code

Codex

npx skills add daocloud/daocloud-skills --skill '*' -g --agent codex

Cursor

npx skills add daocloud/daocloud-skills --skill '*' -g --agent cursor

Use --agent more than once to install into multiple agents at once:

npx skills add daocloud/daocloud-skills --skill '*' -g \
  --agent claude-code \
  --agent codex \
  --agent cursor

This installs from the repository’s default branch. To inspect the available skills (and their names, for --skill) before installing:

npx skills add daocloud/daocloud-skills --list

To uninstall a skill, use remove with its name and the same agent target:

npx skills remove -g dce --agent codex

After installation, open a new agent session and ask it to use the dce skill for DCE operations, for example:

Use the dce skill to list DCE clusters.

Install the CLI

The recommended way is via npm:

npm install -g @daocloud-cli/dce

Or run on demand without a global install:

npx @daocloud-cli/dce --help

The npm package downloads the matching prebuilt binary from GitHub Releases on install and verifies it against checksums.txt. Supported platforms: darwin/linux/windows on amd64/arm64, Node.js >= 16.

Alternatively, download a prebuilt archive from GitHub Releases directly:

VERSION=v0.2.0-rc.4
OS=darwin
ARCH=amd64
PKG="dce-${VERSION}-${OS}-${ARCH}"

curl -fL "https://github.com/DaoCloud/daocloud-skills/releases/download/${VERSION}/${PKG}.tar.gz" -o dce.tar.gz
tar -xzf dce.tar.gz
sudo install -m 0755 "${PKG}/dce" /usr/local/bin/dce

Use OS=darwin for macOS and OS=linux for Linux. Use ARCH=arm64 for Apple Silicon or ARM64 Linux, and ARCH=amd64 for Intel macOS or x86_64 Linux.

On Windows, use PowerShell to download and extract the ZIP archive:

$VERSION = "v0.2.0-rc.4"
$PKG = "dce-${VERSION}-windows-amd64"
Invoke-WebRequest "https://github.com/DaoCloud/daocloud-skills/releases/download/${VERSION}/${PKG}.zip" -OutFile dce.zip
Expand-Archive dce.zip -DestinationPath . -Force
& ".\${PKG}\dce.exe" --help

The npm installation works in Windows PowerShell as well:

npm install -g @daocloud-cli/dce
dce --help

Verify the CLI is available:

dce --help

For source builds from this repository:

make build
sudo cp bin/dce /usr/local/bin/dce

For local development, use make dev to build the CLI, copy it to /usr/local/bin/dce, and symlink the local skill directory:

make dev

Usage

# Log in to a DCE instance
dce auth login --hostname https://

# Browse available commands
dce commands --json

# Search for a command by intent
dce search "list clusters" --json

# Inspect a command before executing
dce commands show container-management cluster list-clusters --json

# Execute a command
dce container-management cluster list-clusters -o json

Container Image

The image bundles the dce binary and the skills/dce directory under /app/, intended for use as a Kubernetes init container to distribute the tooling into a shared volume.

# Build multi-arch image locally
make image

# Build and push to registry
make image-push IMAGE_REPO=registry.example.com/dce IMAGE_TAG=v1.0.0

Default values: IMAGE_REPO=daocloud/dce, IMAGE_TAG=latest.

Init Container Example

initContainers:
  - name: install-dce
    image: daocloud/dce:latest
    command:
      - sh
      - -c
      - |
        cp /app/dce /target/bin/dce
        cp -r /app/skills/dce /target/.agents/skills/dce
    volumeMounts:
      - name: tools
        mountPath: /target

volumes:
  - name: tools
    emptyDir: {}

After the init container completes, the shared volume contains:

• /target/bin/dce — CLI binary
• /target/.agents/skills/dce/ — AI agent skill

Development

Project Structure

.
├── cli.yaml                  # CLI name and auth config
├── specs/sources.yaml        # OpenAPI source definitions (pinned commits)
├── internal/
│   ├── generated/            # Generated Go command modules (do not edit)
│   └── overlay/              # Per-source field overrides for codegen
├── skills/dce/            # AI agent skill (SKILL.md + references)
├── docs/                     # Developer guides
├── cmd/dce/main.go        # CLI entrypoint
└── doc.go                    # Embeds cli.yaml for use by main

Guides

• Adding a new product — step-by-step guide for onboarding a new DCE product (spec source, overlay, codegen, verification)

How It Works

1. specsync clones the pinned commit of daocloud-api-docs and extracts the OpenAPI JSON files into .cache/specs-sync/.
2. codegen reads each spec, applies the overlay from internal/overlay/.yaml, and emits:
   • Go cobra subcommands under internal/generated//
   • A command index at skills/dce/references/modules/.md
   • An updated internal/generated/modules_gen.go that mounts all modules
3. go build compiles everything into a single static binary bin/dce.

Overlay files are the only place where human-maintained configuration lives — everything else is generated and should not be edited by hand.

Updating a Pinned Spec

Update the pinned_tag for the relevant source in specs/sources.yaml, then run:

make sync-one SOURCE=