Agentar

Overview

Use this skill to run Agentar from the command line, delegate specific tasks to Agentar platform agents, and validate Agentar workflows end to end. Ordinary agentar "..." chat does not require a backend agentId; select an explicit --agent only when the task should use one of the built-in scenario agents or when the caller supplies a backend agentId.

Agentar can delegate user tasks to built-in scenario agents: office-agent for PPTs, research documents, summaries, and tables; image-agent for image generation; and video-agent for video creation. This skill provides the safe CLI workflow for running ordinary chat or a selected scenario agent with explicit credentials and verifying results with repeatable commands.

The built-in scenario agents currently exposed by the CLI are:

• office-agent: delegate office productivity tasks such as creating PPTs, drafting research documents, summarizing materials, and producing tables.
• image-agent: delegate image generation tasks.
• video-agent: delegate video creation tasks.

The skill also covers installing and verifying the agentar command, configuring credentials safely, running chat/session/attachment flows, selecting built-in scenario agents when useful, and diagnosing ordinary CLI usage problems before changing source code.

The Agentar CLI distribution channel is an implementation detail of the install workflow below. Follow that workflow when installation or upgrade is required; otherwise start from the usage, safety, or troubleshooting sections that match the user's request.

When To Use

Use this skill when the user asks to:

• Ask Agentar a general question or start an ordinary Agentar chat.
• Run a general Agentar task from the command line when no specialized scenario agent is needed.
• Create office outputs such as PPTs, research documents, summaries, or tables.
• Generate images.
• Create videos.
• Run or continue Agentar conversations from the command line without requiring an agent ID.
• Test attachments, output formats, sessions, or generated results.
• Configure Agentar credentials, authentication state, session state, or isolated test state.
• Install, upgrade, verify, or troubleshoot the Agentar CLI or Agentar Skill package.

Do not use this skill for unrelated Agentar platform source-code changes unless the CLI workflow has already reproduced a defect that requires code-level debugging.

Install Workflow

Check Node.js and npm first. The installer uses npm internally; if npm is unavailable, fix npm instead of switching to pnpm or yarn for Agentar CLI installation:

node -v
npm -v

Confirm the install script is reachable:

curl -I 'https://dtaiagtavtr.antdigital.com/agentar-cli/install.sh'

Prefer the public production install script for user-facing installation:

curl -fsSL 'https://dtaiagtavtr.antdigital.com/agentar-cli/install.sh' | bash

The installer places the binary under /tmp/agentar-cli/bin by default. Verify with the absolute path first:

/tmp/agentar-cli/bin/agentar --version
/tmp/agentar-cli/bin/agentar --help

Add the binary to the current shell only when direct agentar invocation is needed:

export PATH="/tmp/agentar-cli/bin:$PATH"
agentar --version

For validation or CI-style checks, keep token and state files isolated under a temp directory.

The public production skill zip is available at:

https://dtaiagtavtr.antdigital.com/agentar-cli/skill.zip

Install the marketplace Skill with the official Skills command:

npx skills add OpenAgentar/agentar-skills

Usage Workflow

1. Verify the binary first:

   agentar --version
   agentar --help
   

2. Configure credentials. Prefer environment variables for one-off runs and auth login for persistent local use:

   AGENTAR_API_KEY=<YOUR_API_KEY> agentar "hello"
   agentar auth login --api-key <YOUR_API_KEY>
   agentar auth status
   

3. Run chat commands:

   agentar "hello"
   agentar chat "hello"
   agentar "hello" --output json
   

4. Use session commands only when multi-turn continuation is required:

   agentar "remember this" --name demo
   agentar -c "continue"
   agentar session list
   agentar session use demo
   

5. Do not require a backend agentId for ordinary chat. Use --agent only when the task clearly maps to a built-in scenario agent or when the caller explicitly supplies a backend agentId. Prefer built-in --agent presets over manual scene flags; use a raw backend agentId only when the caller explicitly supplies one:

   agentar "hello"
   agentar "summarize this" --attach ./file.pdf
   agentar -a office-agent "draft an office document outline"
   agentar -a image-agent "generate an image"
   agentar -a video-agent "generate a 5 second video of an orange cat resting by a sunny window" --duration 5
   agentar --agent openapi_default_agent "hello"
   

   agentar agents lists local presets only: office-agent, image-agent, and video-agent. Unknown --agent values are passed through as backend agentId values and are not listed by the CLI.

Safety And Secrets

Do not print API keys, bearer tokens, saved token files, or raw request headers. Use placeholders such as <YOUR_API_KEY> or sk-xxxxxxxxxxxX in instructions and reports. For repeatable tests, isolate state with AGENTAR_TOKEN_FILE and AGENTAR_PLAYGROUND_STATE_FILE under a temp directory.

Troubleshooting

Resources

• Public production install script: https://dtaiagtavtr.antdigital.com/agentar-cli/install.sh.
• Public production skill zip: https://dtaiagtavtr.antdigital.com/agentar-cli/skill.zip.
• references/agentar-cli-quick-reference.md: compact command, output, attachment, and scenario examples.