Orchestrate

orchestrate

Claude Code plugin for lionagi's multi-agent orchestration. Nine skills and two agent profiles covering workflow planning, execution, quality gating, code review, debugging, and development methodology.

Prerequisites

• Claude Code CLI
• lionagi >= 0.26.0: pip install lionagi
• Optional: Lion Studio for monitoring: li studio

Install

claude /plugin marketplace add ohdearquant/lionagi
claude /plugin install orchestrate@lionagi

Skills

Skill	Description
orchestrate	Plan and execute multi-agent workflows via li o flow, li o fanout, li play
show	Orchestrate multi-play shows with quality gates and adaptive replanning
playbook	Author .playbook.yaml files — reusable parametric workflow templates
pr-review	Multi-perspective PR review with parallel specialist reviewers and critic synthesis
review	General-purpose code review checklist (correctness, API, tests, readability, security)
security-review	Threat-model security review rubric with CWE mapping and severity calibration
debug	Systematic debugging workflow: research → orchestrate agents → escalate
summarize	Mid-session context capture: checkpoint decisions, patterns, and progress
tdd	Test-driven development discipline: Red-Green-Refactor with gate checks

Agent profiles

Agent	Role
orchestrator	DAG planner: decomposes tasks, assigns workers, manages artifacts, synthesizes results
critic	Quality gate: adversarial review, evidence-based verdicts (APPROVE/REJECT)

Quick start

# Run a playbook
li play feature "add user authentication"

# Fan out parallel workers
li o fanout claude "audit this module for dead code" -n 4

# Plan a DAG flow
li o flow claude "refactor the auth module" --dry-run

# Start Studio for monitoring
li studio

lionagi folder setup

The ~/.lionagi/ directory is created on first use:

~/.lionagi/
├── playbooks/          # .playbook.yaml files (li play reads from here)
├── agents/             # Agent profiles (<name>/<name>.md or <name>.md)
├── runs/               # Run persistence (auto-managed)
├── shows/              # Show workspaces (auto-managed)
├── worktrees/          # Git worktrees for isolated play execution
├── teams/              # Team inbox files (auto-managed)
├── skills/             # CC-compatible skill files
├── settings.yaml       # Global settings (model defaults, hooks)
└── state.db            # SQLite state database (sessions, shows, schedules)

Sample playbooks

Ready-to-use playbooks in examples/playbooks/ in the lionagi repo:

Playbook	Purpose
feature.playbook.yaml	End-to-end feature implementation
pr-review.playbook.yaml	Multi-perspective PR review
test-coverage.playbook.yaml	Iterative test coverage
research.playbook.yaml	Technical research pipeline
resolve-issues.playbook.yaml	GitHub issue resolution
doc-alignment.playbook.yaml	Documentation generation/alignment

Copy any of them to get started:

cp examples/playbooks/feature.playbook.yaml ~/.lionagi/playbooks/
li play feature "add OAuth login"

Getting help

• Issues: https://github.com/ohdearquant/lionagi/issues
• Discussions: https://github.com/ohdearquant/lionagi/discussions
• Source: lionagi/cli/ for CLI, apps/studio/ for Studio, lionagi/state/ for data model

Source code reference

• CLI orchestration: lionagi/cli/orchestrate/flow.py, lionagi/cli/orchestrate/fanout.py
• Agent system: lionagi/cli/agent.py, lionagi/agent/
• Playbooks: ~/.lionagi/playbooks/*.playbook.yaml
• State DB schema: lionagi/state/schema.sql
• Studio: apps/studio/server/, apps/studio/frontend/
• Scheduler: apps/studio/server/scheduler/