Overview

Orchestration is a multi-agent task coordination framework that uses a shared SQLite queue and markdown visibility. Tasks are claimed atomically, can declare dependencies, and support timeouts and retries. The system uses markdown interchange to publish human-readable task cards.

How This Skill Works

Agents register and create tasks in a central SQLite queue (the DB is the source of truth). Tasks are claimed atomically by a single agent, with optional dependencies enforcing order. A sweep handles stale tasks and retries, while refresh generates readable .md task views via @openclaw/interchange.

When to Use It

• Coordinating parallel coding, research, or QA tasks across multiple agents
• Maintaining human-readable task visibility with markdown cards
• Enforcing single-agent claims to prevent duplicate work
• Managing task dependencies and automatic retries on timeout or failure
• Generating up-to-date task dashboards via markdown interchange

Quick Start

1. Step 1: cd skills/orchestration; npm install; node src/cli.js agent register my-agent --capabilities \"coding,research\"
2. Step 2: node src/cli.js task create \"Build feature X\" --desc \"...\" --priority high; node src/cli.js task claim <task-id> --agent my-agent
3. Step 3: node src/cli.js task complete <task-id> --summary \"Done\"; node src/cli.js refresh

Best Practices

• Define clear task descriptions and explicit dependencies to minimize ambiguity
• Treat the SQLite DB as the single source of truth; keep projections read-only
• Use atomic claims and meaningful timeouts to guard against race conditions
• Run regular sweeps to reassign or retry stale tasks promptly
• Use refresh to keep .md task cards synchronized for visibility

Example Use Cases

• Build feature X by coordinating frontend, backend, and tests with dependency order
• Data pipeline: fetch -> transform -> load, with retries on transient failures
• QA regression tasks sequenced to ensure prerequisites complete before tests
• Literature-backed research tasks where findings unlock subsequent steps
• Code review and feature flag rollout tasks that require prior approvals

Frequently Asked Questions