# {{name}}

{{description}}

> AI / ML Engineering harness · domain: `ai/ml-lifecycle`. Generated with [create-agent-harness](https://github.com/ruvnet/agent-harness-generator).

## Behavioral rules

- Use the harness's MCP tools (`mcp__{{name}}__*`) for orchestration
- Memory and routing are handled by the kernel — you don't need to learn them
- Defer destructive operations to the user

## Agents

| Agent | Tier | Role |
|---|---|---|
| `data-curator` | sonnet | Builds and documents the dataset. |
| `trainer` | sonnet | Runs reproducible training jobs. |
| `evaluator` | opus | The honest eval gate. |
| `deployer` | sonnet | Ships behind a guardrail. |
## Skills

- `/eval-report` — Produce an honest eval report: metrics, subgroup slices, baseline delta, ship/no-ship.

## Commands

- `doctor` — Health-check the harness: kernel load, MCP wiring, memory backend, host adapter.

## Architecture

This harness uses [@metaharness/kernel](https://www.npmjs.com/package/@metaharness/kernel) — a Rust-compiled WASM module with a NAPI-RS native fallback — so the same code runs identically on every platform.