substrate-selection

What problem does it solve?

It prevents orchestration and adapter layers from being tangled with unverified runtime choices, reducing workflow breakage when swapping coding substrates or providers.

Core Features & Use Cases

• Role split for reliability: Cleanly separates orchestrator substrate, leaf executor, and skill surface so policies don’t accidentally become runtime invocations.
• Grounded selection policy: Chooses defaults only from verified docs or local runs, and labels indirect evidence as provisional rather than fact.
• Provider vs harness delta handling: Distinguishes model/provider differences from harness/runtime guarantees, then rebuilds missing harness controls in skills/config.
• Compensating controls for weaker stacks: Adds structured response packets, bounded retries/timeouts, observable progress signals, and escalation for stalled runs.
• Integration rule for inspectability: Ensures each run records orchestrator, delegated external-harness, and leaf-agent lanes with concrete endpoint/model bindings.

Quick Start

In your agentic coding harness, apply the substrate-selection matrix to pick orchestrator, delegated harness, and leaf editor substrates, then encode the chosen lanes and compensating controls explicitly in your run contract.