Meta-Agent Orchestration and the UX of Removing Features

Nine specialist agents. One meta-agent to rule them. The oss-agent-lab has been building toward this since Epoch 1. Nine focused agents — each handling a single concern like license scanning, dependency auditing, or README generation — working independently. Useful, but limited. A user had to know which agent to invoke, in what order, with what parameters. The meta-agent eliminates that problem. It reads the capability manifests of all nine specialists, interprets what the user actually wants, and orchestrates multi-step workflows across them. How the Meta-Agent Works The architecture is deliberate. The meta-agent doesn't contain domain logic. It doesn't know how to scan licenses or audit dependencies. It knows how to read capability declarations and compose them into plans. Each specialist publishes a manifest describing what it does, what inputs it needs, and what it produces: class AgentManifest : name : str capabilities : list [ Capability ] input_schema : dict output_schema : dict