Alibabacloud Workbench CLI

Workbench CLI Expert

Help users operate Alibaba Cloud ECS instances — especially those without public IP addresses — using the workbench command-line tool. Core capabilities for Agent workflows: millisecond-level remote command execution (exec), file transfer up to 1GB (upload/download), and port forwarding. This skill covers: install → configure credentials → exec/transfer/forward → manage sessions → troubleshoot errors.

Instructions

1. Install the Workbench CLI

Pre-check:

workbench version     # Should print version, commit, build date

Linux / macOS:

curl -fsSL https://workbench-cli.oss-cn-hangzhou.aliyuncs.com/install.sh | bash

Windows (PowerShell):

irm https://workbench-cli.oss-cn-hangzhou.aliyuncs.com/install.ps1 | iex

Upgrade:

workbench upgrade                        # Upgrade to latest
workbench upgrade --version 0.2.0        # Upgrade to specific version

2. Configure credentials

Credentials are stored in ~/.workbench/config.json with 0600 permissions. As an Agent, write this file directly instead of using the interactive workbench config command.

# Create config directory
mkdir -p ~/.workbench

# Write config file (example: AK mode)
cat > ~/.workbench/config.json << 'EOF'
{
  "current": "default",
  "profiles": {
    "default": {
      "mode": "AK",
      "access_key_id": "<AccessKeyID>",
      "access_key_secret": "<AccessKeySecret>"
    }
  }
}
EOF

# Set secure permissions
chmod 600 ~/.workbench/config.json

Config file schema by mode:

AK mode:

{
  "current": "default",
  "profiles": {
    "default": {
      "mode": "AK",
      "access_key_id": "LTAI...",
      "access_key_secret": "..."
    }
  }
}

RamRoleArn mode (auto-refreshes STS tokens):

{
  "current": "default",
  "profiles": {
    "default": {
      "mode": "RamRoleArn",
      "access_key_id": "LTAI...",
      "access_key_secret": "...",
      "ram_role_arn": "acs:ram::123456789:role/WorkbenchRole",
      "role_session_name": "workbench-session"
    }
  }
}

CredentialsURI mode (HTTP endpoint returns credentials):

{
  "current": "default",
  "profiles": {
    "default": {
      "mode": "CredentialsURI",
      "credentials_uri": "http://localhost:8080/credentials"
    }
  }
}

Profile management (non-interactive):

workbench config list                     # List all profiles (* marks active)
workbench config switch --profile prod    # Switch active profile
workbench config get                      # Show current profile details (JSON)
workbench config get --profile prod       # Show specific profile details
workbench config delete --profile old     # Delete a profile (cannot delete active)

3. Command reference

workbench
├── exec             # Execute remote command (non-interactive, millisecond-level)
├── upload           # Upload local file to instance (up to 1GB, via OSS relay)
├── download         # Download file from instance (up to 1GB, via OSS relay)
├── list             # List ECS instances
├── session          # Session management (list / close)
├── daemon           # Daemon lifecycle (start / status / stop)
├── config           # Credential configuration & profile management
│   ├── set          # Set a single config field
│   ├── list         # List all profiles
│   ├── switch       # Switch active profile
│   ├── get          # Show profile details
│   └── delete       # Delete a profile
├── upgrade          # Self-update
└── version          # Print version info

Global flags:

4. List instances

workbench list ecs --region cn-hangzhou
workbench list ecs --region cn-hangzhou --status Running
workbench list ecs --region cn-hangzhou --tag env=prod --tag team=infra
workbench list ecs --region cn-hangzhou --instance-type ecs.g7.large
workbench list ecs --region cn-hangzhou --instance-name my-instance
workbench list ecs --region cn-hangzhou --image-id ubuntu_22_04_x64_20G_alibase_20230907.vhd
workbench list ecs --region cn-hangzhou --output json

--region is required for list ecs. Filters: --status (Running|Stopped|Starting|Stopping), --tag (key=value or key, repeatable, AND logic), --instance-type (e.g. ecs.g7.large), --instance-name (supports wildcards *), --image-id, --vpc-id, --zone-id, --vswitch-id, --private-ip (comma-separated), --limit (1-100, default 50), --next-token (pagination).

JSON output schema:

[{
  "instance_id": "i-bp1xxxxx",
  "instance_name": "web-prod-01",
  "instance_type": "ecs.g7.large",
  "region_id": "cn-hangzhou",
  "status": "Running",
  "private_ip": "172.16.0.10",
  "public_ip": "",
  "os_type": "linux",
  "image_id": "ubuntu_22_04_x64_20G_alibase_20230907.vhd",
  "tags": {"env": "prod"}
}]

5. Remote command execution

Built-in safety: The CLI is a non-interactive executor — it does NOT provide a persistent shell. Each invocation is isolated and stateless, which prevents accidental cascading damage from lingering shell sessions.

Agent pre-check requirement: Before executing destructive commands (e.g., rm -rf, shutdown, reboot, mkfs, dd, service stop/restart, or any command that deletes data or halts the system), the Agent MUST confirm with the user by describing the intended action, the target instance, and the potential impact. Do NOT execute destructive commands without explicit user approval.

workbench exec --instance-id i-bp1xxxxx --command "df -h"
workbench exec --instance-id i-bp1xxxxx --command "sleep 30" --timeout 10
workbench exec --instance-id i-bp1xxxxx --command "df -h" --output json

Important: Each exec invocation runs in an independent shell context. State (cd, export) is NOT preserved between calls. Use && or ; to chain commands that need shared context in a single invocation.

JSON output schema:

{
  "output": "Filesystem ...\n",
  "stderr": "",
  "exit_code": 0
}

6. File transfer

Built-in safety: The upload command has a built-in overwrite protection. When the remote destination file already exists, the CLI prompts for confirmation before overwriting:

Remote file "/root/id.txt" already exists (728 B, modified Jul 27 10:42). Overwrite? [y/N]

The default answer is No — if the user does not explicitly confirm, the upload is aborted. This prevents accidental overwriting of existing remote files.

Agent pre-check requirement: When using upload in automated/Agent workflows where interactive confirmation is not possible, the Agent MUST first check whether the target file exists on the remote instance (e.g., via workbench exec --command "ls -la <path>") and inform the user if a file will be overwritten.

workbench upload ./app.jar /opt/app/app.jar --instance-id i-bp1xxxxx
workbench download /var/log/app.log ./ --instance-id i-bp1xxxxx