# Sprint 157 Plan: Real Multi-Language MCP Execution for Sync/Regeneration/Merge

## Context

This sprint replaces scaffolded MCP tool handlers with concrete runtime logic
for representative languages, starting with sync/regeneration/merge surfaces.

---

## Goals

1. Replace canned tool responses with language-aware runtime execution
2. Add deterministic argument validation and normalization
3. Persist structured checkpoints/traces for replay and diagnostics
4. Keep backward-compatible response fields for existing clients/tests

---

## Steps

### Step 1809: Shared multi-language runtime state model for MCP handlers (10 tests)
### Step 1810: Checkpoint save/restore/replay runtime implementation model (10 tests)
### Step 1811: Generic sync execution response model backed by profile store (10 tests)
### Step 1812: Generic regeneration execution response model backed by policy model (10 tests)
### Step 1813: `whetstone_save_hybrid_checkpoint` runtime implementation (8 tests)
### Step 1814: `whetstone_restore_hybrid_checkpoint` runtime implementation (8 tests)
### Step 1815: `whetstone_replay_hybrid_session` runtime implementation (8 tests)
### Step 1816: Profile-backed readiness and qualification packet model (8 tests)
### Step 1817: Runtime execution trace report artifact for sync/regeneration/merge tools (8 tests)
### Step 1818: Sprint 157 integration summary + regression (8 tests)

---

## Architecture Gate

- Runtime state updates must be deterministic under identical inputs
- Handler responses must include profile-derived fields, not hardcoded placeholders
- Existing compatibility fields (`success`, `tool`, `status`, `data`) must remain
- Max 600 lines per header