# Sprint 158 Plan: Real Multi-Language Constructive Loop and Toolchain Routing

## Context

After sync/regeneration/merge runtimeization, Sprint 158 implements real
toolchain/provider and constructive loop behavior for representative languages.

---

## Goals

1. Back toolchain provider tools with real language/provider mappings
2. Implement constructive step/status/loop with explicit stage evidence
3. Normalize diagnostics with source, severity, location, and fix hints
4. Preserve deterministic fallback policy when providers are unavailable

---

## Steps

### Step 1819: Representative language toolchain provider registry model (10 tests)
### Step 1820: Provider probing and fallback routing implementation model (10 tests)
### Step 1821: Diagnostic normalization runtime implementation model (10 tests)
### Step 1822: Constructive stage execution ledger model (10 tests)
### Step 1823: `whetstone_list_toolchain_providers` runtime implementation (8 tests)
### Step 1824: `whetstone_probe_toolchain_provider` runtime implementation (8 tests)
### Step 1825: `whetstone_normalize_diagnostics` runtime implementation (8 tests)
### Step 1826: `whetstone_run_constructive_step`/`status`/`loop` runtime packet model (8 tests)
### Step 1827: Cross-language constructive execution report artifact (8 tests)
### Step 1828: Sprint 158 integration summary + regression (8 tests)

---

## Architecture Gate

- Constructive stage transitions must be explicit and replayable
- Provider fallback must emit structured rationale and selected route
- Diagnostics normalization must preserve provenance to raw provider output
- Max 600 lines per header