Skill Detail

Run declarative multi-agent CLI workflows with Comanda

Turns repeatable AI work into versioned YAML pipelines that can call Claude Code, Codex, Gemini CLI, Kimi Code, API models, tools, files, worktrees, and MCP-exposed workflows from the terminal.

Developer ToolsMulti-Framework
Developer Tools Multi-Framework Security Reviewed
⭐ 315 GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill run-declarative-multi-agent-cli-workflows-with-comanda Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
Comanda CLI, configured model or agent providers such as Claude Code, Codex, Gemini CLI, Kimi Code, or API models, optional MCP client/server setup
Install & setup
Install with brew install kris-hansen/comanda/comanda or go install github.com/kris-hansen/comanda@latest, run comanda configure, generate a workflow with comanda generate workflow.yaml "review this code for bugs", then run it with comanda process workflow.yaml.
Author
kris-hansen
Publisher
Independent
Last updated
Jul 27, 2026
Quick brief

## What the agent does
The agent uses Comanda to generate, inspect, improve, and run YAML workflow files. It can pipe files or command output through a workflow, orchestrate multi-model and multi-agent steps, process batches, call allowlisted tools, manage codebase context, and expose discovered workflows as MCP tools.

How it works

What this skill actually does

## When to use it
Use this when an operator has a repeatable AI task that should be version controlled, reviewed, shared, rerun, or exposed to other agent clients instead of being typed as an ad hoc prompt each time. Good fits include review pipelines, batch file processing, agentic loops, worktree workflows, and command-line automation with explicit tool boundaries.

## Scope boundary
This is not a generic CLI, framework, or model wrapper listing. The skill boundary is declarative operator workflows: a YAML file defines inputs, model or agent steps, tool permissions, outputs, and optional MCP exposure, and the agent runs or improves that workflow.

## Source evidence
The upstream repository documents Homebrew and Go installation, comanda configure, generate, process, chart, and improve commands, minimal YAML workflow syntax, multi-agent examples, agentic-loop examples, tool allowlists, worktree workflows, and MCP server mode.