WIP: stage all uncommitted work — sprints 46-221, graduation headers, specialist fleet, test steps 909-1988

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

sprint58_plan.md

@@ -0,0 +1,36 @@
+# Sprint 58 Plan: Legacy Ingestion v2 (Semantic Recovery Engine)
+
+## Context
+
+AST-first projection is easy; ingestion remains the bottleneck. Sprint 58 upgrades
+legacy ingestion into a semantics-recovery pipeline feeding the new IR.
+
+---
+
+## Goals
+
+1. Recover missing intent from legacy code with confidence scoring
+2. Surface ambiguity packets and review tasks automatically
+3. Add multi-file/system-level recovery (APIs, build, runtime assumptions)
+4. Produce migration-ready IR even from weakly annotated legacy sources
+
+---
+
+## Steps
+
+### Step 809: Legacy semantic recovery graph engine (12 tests)
+### Step 810: Cross-file API intent inference (10 tests)
+### Step 811: Build/runtime assumption inference (10 tests)
+### Step 812: Ambiguity packet model (`assumed`, `unknown`, `conflict`) (10 tests)
+### Step 813: Human review queue generator from ambiguities (10 tests)
+### Step 814: Confidence calibration for recovered intent (10 tests)
+### Step 815: Legacy->IR migration readiness scoring (8 tests)
+### Step 816: `whetstone_ingest_legacy_to_ir` MCP tool (8 tests)
+### Step 817: Legacy modernization dossier generator (8 tests)
+### Step 818: Sprint 58 integration summary + regression (8 tests)
+
+---
+
+## Quality Target
+
+- Reduce manual ambiguity resolution load while keeping fail-closed behavior.