diff --git a/tests/test_full_stack_contract.py b/tests/test_full_stack_contract.py
index 59cee34..6ab4a91 100644
--- a/tests/test_full_stack_contract.py
+++ b/tests/test_full_stack_contract.py
@@ -1,10 +1,32 @@
+import asyncio
+import base64
+import json
 from pathlib import Path
 
+from a2a_pack import LocalRunContext, PlatformUserAuth
 from a2a_pack.cli.local import load_local_project
 
+from agent import BrowserDocument, QuoteInput, QuoteJudgeStudioV1, QuoteWeights, _parse_browser_documents
 
 ROOT = Path(__file__).resolve().parents[1]
 
+SUCCESS_ARGS = {
+    "comparison_id": "studio-quote-v1",
+    "quotes": [
+        {"vendor": "Acme Bearings", "unit_price": 10.75, "quantity": 500, "delivery_days": 12, "warranty_months": 12},
+        {"vendor": "Beta Industrial", "unit_price": 9.1, "quantity": 500, "delivery_days": 8, "warranty_months": 18},
+    ],
+    "weights": {"price": 50, "delivery": 30, "warranty": 20},
+}
+
+
+def _ctx():
+    return LocalRunContext(
+        auth=PlatformUserAuth(sub="test-user", user_id=42, email="quote@example.test"),
+        task_id="test-task",
+        caller="pytest",
+    )
+
 
 def test_full_stack_product_contract():
     manifest = (ROOT / "a2a.yaml").read_text(encoding="utf-8")
@@ -12,17 +34,87 @@ def test_full_stack_product_contract():
     frontend = (ROOT / "frontend" / "src" / "App.jsx").read_text(encoding="utf-8")
 
     assert "frontend:" in manifest and "mount: /app" in manifest
+    assert "public: false" in manifest
     assert "resources:" in manifest and "databases:" in manifest
     assert "migrations:" in manifest and "db/migrations" in manifest
     assert "PlatformUserAuth" in source
     assert "AgentPlatformResources" in source and "AgentDatabase(" in source
-    assert "Skill runner" not in frontend, "replace the generic scaffold with the product workflow"
+    assert "LLMProvisioning" not in source and "ctx.llm" not in source
+    assert "Skill runner" not in frontend
+    assert "QuoteJudge Studio" in frontend
+    assert "config.json" not in frontend, "frontend should use a2a.js config loader rather than hard-code platform metadata"
+    assert "A2A_LITELLM" not in frontend and "DATABASE_URL" not in frontend
 
     migrations = list((ROOT / "db" / "migrations").glob("*.sql"))
     assert migrations and all(path.read_text(encoding="utf-8").strip() for path in migrations)
 
     card = load_local_project(ROOT).agent_cls().card()
+    skill_names = {skill.name for skill in card.skills}
+    assert {"compare_quotes", "get_comparison", "upload_quotes", "upload_quote_file"}.issubset(skill_names)
     databases = card.runtime.platform_resources.databases
     assert databases, "live Agent Card must declare its managed database"
     assert databases[0].migrations is not None
     assert databases[0].migrations.path == "db/migrations"
+    assert card.runtime.pricing.caller_pays_llm is False
+
+
+def test_compare_reload_failure_and_receipt():
+    agent = QuoteJudgeStudioV1()
+    ctx = _ctx()
+
+    result = asyncio.run(agent.compare_quotes(
+        ctx,
+        SUCCESS_ARGS["comparison_id"],
+        [QuoteInput.model_validate(item) for item in SUCCESS_ARGS["quotes"]],
+        QuoteWeights.model_validate(SUCCESS_ARGS["weights"]),
+    ))
+    assert result["ok"] is True
+    assert result["comparison_id"] == "studio-quote-v1"
+    assert result["recommendation"]["vendor"] == "Beta Industrial"
+    assert result["receipt"]["persisted"] is True
+    assert "receipt_id" in result["receipt"]
+
+    reopened = asyncio.run(agent.get_comparison(ctx, "studio-quote-v1"))
+    assert reopened["ok"] is True
+    assert reopened["recommendation"]["vendor"] == "Beta Industrial"
+
+    failure = asyncio.run(agent.compare_quotes(
+        ctx,
+        "studio-quote-invalid",
+        [QuoteInput(vendor="Only Vendor", unit_price=10, quantity=1, delivery_days=5, warranty_months=6)],
+        QuoteWeights(price=50, delivery=30, warranty=20),
+    ))
+    assert failure["ok"] is False
+    assert failure["code"] == "at_least_two_quotes_required"
+    assert "action" in failure
+
+
+def test_browser_upload_bridge_bounds_and_parses():
+    payload = json.dumps({"quotes": SUCCESS_ARGS["quotes"]}).encode("utf-8")
+    parsed = _parse_browser_documents([
+        BrowserDocument(filename="quotes.json", media_type="application/json", data_base64=base64.b64encode(payload).decode("ascii"))
+    ])
+    assert parsed["ok"] is True
+    assert len(parsed["quotes"]) == 2
+
+    too_large = base64.b64encode(b"x" * (64 * 1024 + 1)).decode("ascii")
+    invalid = _parse_browser_documents([
+        BrowserDocument(filename="too-big.txt", media_type="text/plain", data_base64=too_large)
+    ])
+    assert invalid["ok"] is False
+    assert invalid["code"] == "upload_too_large"
+
+    bad_b64 = _parse_browser_documents([
+        BrowserDocument(filename="bad.csv", media_type="text/csv", data_base64="not valid base64")
+    ])
+    assert bad_b64["ok"] is False
+    assert bad_b64["code"] == "invalid_base64"
+
+
+def test_mcp_typed_tool_schemas_are_present():
+    card = QuoteJudgeStudioV1().card()
+    schemas = {skill.name: skill.input_schema for skill in card.skills}
+    compare_schema = schemas["compare_quotes"]
+    assert set(compare_schema["required"]) == {"comparison_id", "quotes", "weights"}
+    assert "quotes" in compare_schema["properties"]
+    assert schemas["upload_quote_file"]["properties"]["file"]["x-a2a-file-upload"]["max_bytes"] == 64 * 1024