
from swarm-cli15
Orchestrate multi-agent AI workflows using swarm-cli compose (swarm.yaml) files with planner/doer/reviewer pipelines and DAG task dependencies.
Swarm CLI's 'swarm up' skill defines and runs multi-agent workflows using a swarm.yaml compose format. It provides a planner/doer/reviewer pattern, task lifecycle conventions (.todo.md → .processing.md → .done.md), and pipeline configuration for iterations and parallelism. The tool manages agent lifecycle (run, pause, kill, inspect) and supports runtime variables for inter-task communication.
Use this skill when you need to coordinate multiple AI agents or automated workers in DAG pipelines: creating swarm compose files, defining task dependencies and runtime variables, running pipelines in foreground or detached mode, monitoring agents, and debugging concurrent workflows. It's suitable for iterative automation, CI-style agent loops, and orchestrated prompt-driven task execution.
Works with agents and runtimes that can run CLI tools and manage file-based task state. Commonly used with local CLIs, containerized runtimes, and prompt-driven agent runners.
Documentation-only skill for swarm-cli, a multi-agent AI workflow orchestrator. SKILL.md provides clear YAML reference for swarm.yaml compose files with planner/doer/reviewer patterns and DAG dependencies. No bundled scripts to test. Well-organized with good examples and reference tables, though some steps could be more specific. Niche tool requiring separate CLI installation.
Clean documentation skill with no security concerns. The skill is essentially a well-written man page for swarm-cli compose files. No executable content to audit beyond the SKILL.md itself.