Skip to content

cli_agent.md

Latest commit

 

History

History
94 lines (65 loc) · 2.96 KB

cli_agent.md

File metadata and controls

94 lines (65 loc) · 2.96 KB

Hugind Agent Command Manual

NAME

hugind agent - run and manage agents.

SYNOPSIS

hugind agent <subcommand> [options]

DESCRIPTION

The agent command runs agent definitions or workflows, supports listing, installing, and removing agents, and can orchestrate multi-agent teams.

SUBCOMMANDS

hugind agent run <path> [--cwd <path>] [--log-file <path>] [-- <args...>]

Runs an agent or workflow from a file path, passing any additional arguments through to the agent runtime.

<path> can be:

  • a local agent directory (contains agent.yaml)
  • a direct local entry path
  • a local workflow .yaml file
  • an installed agent name (resolved from the agents install directory)

If the agent has mode: agentic in its agent.yaml, the runtime drives an LLM tool-use loop automatically. Pass the task via --prompt:

hugind agent run agent/ma-reader --prompt "Find all Python files in this project"

Options:

  • --cwd <path>: override runtime working directory and host filesystem root for this run. JS/WASM module loading still stays scoped to the agent folder. If --cwd points outside the agent root, set permissions.filesystem.allow_outside_agent_root: true in agent.yaml.
  • --log-file <path>: write runtime audit logs to an explicit file path.

hugind agent team <goal> --agents <paths> [--backend <config>] [--concurrency <n>]

Orchestrates a multi-agent team to accomplish a goal. A coordinator agent decomposes the goal into tasks, assigns them to team members, and synthesizes the final result.

hugind agent team "Build a REST API for user management" \
  --agents agent/ma-architect,agent/ma-developer,agent/ma-tester,agent/ma-reviewer \
  --backend qwen-32b

Options:

  • --agents <paths>: comma-separated list of agent directories
  • --backend <config>: config name for the coordinator LLM (default: uses default server)
  • --concurrency <n>: max concurrent agents (default: 4)

The team command:

  1. Sends the goal to the coordinator LLM for task decomposition
  2. Parses the returned JSON task array with dependencies
  3. Executes tasks in parallel (respecting dependency order)
  4. Each agent gets shared memory, messaging, and the user's working directory
  5. Synthesizes a final summary after all tasks complete

All agents share a fresh session on the LLM server for KV cache reuse.

hugind agent list

Lists agents installed under ~/.hugind/agents.

hugind agent install <path>

Installs an agent from a local folder, URL, or zip file. Prints requested permissions and asks for confirmation.

Accepted inputs:

  • Local folder containing agent.yaml
  • Direct path to agent.yaml
  • Local .zip containing a single agent
  • Web URL pointing at a folder, agent.yaml, or .zip
  • GitHub github.com/<owner>/<repo>/tree/... URLs

hugind agent remove <name>

Removes an installed agent.

ENVIRONMENT

  • HUGIND_TRACE=1: enable detailed trace output for agentic execution

HELP

Run hugind agent --help to see flags and options.