diff --git a/docs/adaptive_allocation_control_plane.zh-CN.md b/docs/adaptive_allocation_control_plane.zh-CN.md
index 8adb30d..39c7bc9 100644
--- a/docs/adaptive_allocation_control_plane.zh-CN.md
+++ b/docs/adaptive_allocation_control_plane.zh-CN.md
@@ -14,6 +14,19 @@ P0 提供统一、只读的 Shadow 决策记录，不提供交易授权。它解
 输出为 `qsl.selection_decision.v1`，完整保存候选、拒绝原因、平台选择、策略分数和
 输入摘要。输出固定为 `authority=shadow_only`、`no_order=true` 且所有建议权重为零。
 
+## 通用接入边界
+
+任何策略仓库或平台仓库都可以向 `quant-adaptive-selection` 提交
+`qsl.selection_input.v1` JSON，并得到可保存、可回放的决策工件：
+
+```bash
+quant-adaptive-selection --input selection-input.json --output selection-decision.json
+```
+
+输入必须提供带时区的平台健康快照、版本化市场上下文、不可变候选 release、插件风险
+缩放和冻结策略。命令不接受 broker 凭据、运行时目标或下单参数；输出文件仅是 JSON
+工件，不会修改平台或调度器。
+
 ## 固定边界
 
 - 不读取新闻叙事并直接交易；因子必须来自版本化的数据链。
diff --git a/pyproject.toml b/pyproject.toml
index d9d79ec..90945aa 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -27,6 +27,7 @@ dev = [
 [project.scripts]
 quant-lifecycle = "quant_platform_kit.strategy_lifecycle.cli:main"
 quant-strategy-spec = "quant_platform_kit.strategy_spec.cli:main"
+quant-adaptive-selection = "quant_platform_kit.adaptive_allocation.cli:main"
 
 [tool.setuptools]
 package-dir = { "" = "src" }
diff --git a/src/quant_platform_kit/adaptive_allocation/__init__.py b/src/quant_platform_kit/adaptive_allocation/__init__.py
index 1f8ed06..b65baac 100644
--- a/src/quant_platform_kit/adaptive_allocation/__init__.py
+++ b/src/quant_platform_kit/adaptive_allocation/__init__.py
@@ -14,6 +14,11 @@
     StrategyCandidate,
 )
 from quant_platform_kit.adaptive_allocation.selector import select_shadow
+from quant_platform_kit.adaptive_allocation.io import (
+    SELECTION_INPUT_SCHEMA,
+    build_shadow_selection,
+    load_shadow_selection_input,
+)
 
 __all__ = [
     "AdaptiveSelectionPolicy",
@@ -22,5 +27,8 @@
     "PluginRiskAdjustment",
     "SelectionDecision",
     "StrategyCandidate",
+    "SELECTION_INPUT_SCHEMA",
+    "build_shadow_selection",
+    "load_shadow_selection_input",
     "select_shadow",
 ]
diff --git a/src/quant_platform_kit/adaptive_allocation/cli.py b/src/quant_platform_kit/adaptive_allocation/cli.py
new file mode 100644
index 0000000..46191e5
--- /dev/null
+++ b/src/quant_platform_kit/adaptive_allocation/cli.py
@@ -0,0 +1,27 @@
+"""Command-line entry point for writing a Shadow-only selection record."""
+
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+from typing import Sequence
+
+from quant_platform_kit.adaptive_allocation.io import build_shadow_selection, load_shadow_selection_input
+
+
+def main(argv: Sequence[str] | None = None) -> int:
+    parser = argparse.ArgumentParser(description="Generate a no-order adaptive Shadow selection record")
+    parser.add_argument("--input", required=True, help="versioned selection-input JSON file")
+    parser.add_argument("--output", required=True, help="destination JSON artifact path")
+    args = parser.parse_args(argv)
+
+    decision = build_shadow_selection(load_shadow_selection_input(args.input))
+    output = Path(args.output)
+    output.parent.mkdir(parents=True, exist_ok=True)
+    output.write_text(json.dumps(decision.to_dict(), indent=2, sort_keys=True) + "\n", encoding="utf-8")
+    return 0
+
+
+if __name__ == "__main__":  # pragma: no cover
+    raise SystemExit(main())
diff --git a/src/quant_platform_kit/adaptive_allocation/io.py b/src/quant_platform_kit/adaptive_allocation/io.py
new file mode 100644
index 0000000..ab902d2
--- /dev/null
+++ b/src/quant_platform_kit/adaptive_allocation/io.py
@@ -0,0 +1,196 @@
+"""Strict JSON input/output boundary for Shadow-only adaptive selection."""
+
+from __future__ import annotations
+
+import json
+from collections.abc import Mapping, Sequence
+from datetime import date, datetime
+from pathlib import Path
+from typing import Any
+
+from quant_platform_kit.adaptive_allocation.contracts import (
+    PLATFORM_HEALTH_SCHEMA,
+    MARKET_CONTEXT_SCHEMA,
+    AdaptiveSelectionPolicy,
+    MarketContextSnapshot,
+    PlatformHealthSnapshot,
+    PluginRiskAdjustment,
+    SelectionDecision,
+    StrategyCandidate,
+)
+from quant_platform_kit.adaptive_allocation.selector import select_shadow
+
+
+SELECTION_INPUT_SCHEMA = "qsl.selection_input.v1"
+
+
+def _mapping(value: object, field_name: str) -> Mapping[str, Any]:
+    if not isinstance(value, Mapping):
+        raise ValueError(f"{field_name} must be an object")
+    return value
+
+
+def _sequence(value: object, field_name: str) -> Sequence[object]:
+    if isinstance(value, (str, bytes, bytearray)) or not isinstance(value, Sequence):
+        raise ValueError(f"{field_name} must be an array")
+    return value
+
+
+def _required(mapping: Mapping[str, Any], field_name: str) -> Any:
+    if field_name not in mapping:
+        raise ValueError(f"{field_name} is required")
+    return mapping[field_name]
+
+
+def _bool(value: object, field_name: str) -> bool:
+    if not isinstance(value, bool):
+        raise ValueError(f"{field_name} must be boolean")
+    return value
+
+
+def _number(value: object, field_name: str) -> float:
+    if isinstance(value, bool) or not isinstance(value, (int, float)):
+        raise ValueError(f"{field_name} must be numeric")
+    return float(value)
+
+
+def _integer(value: object, field_name: str) -> int:
+    normalized = _number(value, field_name)
+    if not normalized.is_integer():
+        raise ValueError(f"{field_name} must be an integer")
+    return int(normalized)
+
+
+def _string(value: object, field_name: str) -> str:
+    if not isinstance(value, str):
+        raise ValueError(f"{field_name} must be a string")
+    return value
+
+
+def _date(value: object, field_name: str) -> date:
+    try:
+        return date.fromisoformat(_string(value, field_name))
+    except ValueError as exc:
+        raise ValueError(f"{field_name} must be an ISO date") from exc
+
+
+def _datetime(value: object, field_name: str) -> datetime:
+    try:
+        parsed = datetime.fromisoformat(_string(value, field_name).replace("Z", "+00:00"))
+    except ValueError as exc:
+        raise ValueError(f"{field_name} must be an ISO datetime") from exc
+    if parsed.tzinfo is None or parsed.utcoffset() is None:
+        raise ValueError(f"{field_name} must include a timezone")
+    return parsed
+
+
+def _numeric_mapping(value: object, field_name: str) -> dict[str, float]:
+    return {
+        _string(key, f"{field_name} key"): _number(item, f"{field_name}.{key}")
+        for key, item in _mapping(value, field_name).items()
+    }
+
+
+def _strings(value: object, field_name: str) -> tuple[str, ...]:
+    return tuple(_string(item, f"{field_name} item") for item in _sequence(value, field_name))
+
+
+def _parse_market_context(payload: object) -> MarketContextSnapshot:
+    item = _mapping(payload, "market_context")
+    if item.get("schema") != MARKET_CONTEXT_SCHEMA:
+        raise ValueError("market_context schema is unsupported")
+    return MarketContextSnapshot(
+        as_of=_date(_required(item, "as_of"), "market_context.as_of"),
+        domain=_string(_required(item, "domain"), "market_context.domain"),
+        data_version=_string(_required(item, "data_version"), "market_context.data_version"),
+        data_freshness_days=_integer(_required(item, "data_freshness_days"), "market_context.data_freshness_days"),
+        regime=_string(_required(item, "regime"), "market_context.regime"),
+        regime_confidence=_number(_required(item, "regime_confidence"), "market_context.regime_confidence"),
+        factors=_numeric_mapping(item.get("factors", {}), "market_context.factors"),
+    )
+
+
+def _parse_candidate(payload: object) -> StrategyCandidate:
+    item = _mapping(payload, "candidate")
+    return StrategyCandidate(
+        strategy_profile=_string(_required(item, "strategy_profile"), "candidate.strategy_profile"),
+        release_digest=_string(_required(item, "release_digest"), "candidate.release_digest"),
+        lifecycle_stage=_string(_required(item, "lifecycle_stage"), "candidate.lifecycle_stage"),
+        approved_for_shadow=_bool(_required(item, "approved_for_shadow"), "candidate.approved_for_shadow"),
+        base_score=_number(_required(item, "base_score"), "candidate.base_score"),
+        estimated_volatility=_number(_required(item, "estimated_volatility"), "candidate.estimated_volatility"),
+        factor_exposures=_numeric_mapping(item.get("factor_exposures", {}), "candidate.factor_exposures"),
+        required_plugins=_strings(item.get("required_plugins", []), "candidate.required_plugins"),
+        allowed_platform_ids=_strings(item.get("allowed_platform_ids", []), "candidate.allowed_platform_ids"),
+    )
+
+
+def _parse_platform_health(payload: object) -> PlatformHealthSnapshot:
+    item = _mapping(payload, "platform_health item")
+    if item.get("schema") != PLATFORM_HEALTH_SCHEMA:
+        raise ValueError("platform_health schema is unsupported")
+    return PlatformHealthSnapshot(
+        platform_id=_string(_required(item, "platform_id"), "platform_health.platform_id"),
+        observed_at=_datetime(_required(item, "observed_at"), "platform_health.observed_at"),
+        healthy=_bool(_required(item, "healthy"), "platform_health.healthy"),
+        shadow_capable=_bool(_required(item, "shadow_capable"), "platform_health.shadow_capable"),
+        reconciliation_ok=_bool(_required(item, "reconciliation_ok"), "platform_health.reconciliation_ok"),
+        capacity_score=_number(_required(item, "capacity_score"), "platform_health.capacity_score"),
+        expected_cost_bps=_number(_required(item, "expected_cost_bps"), "platform_health.expected_cost_bps"),
+    )
+
+
+def _parse_plugin_adjustment(payload: object) -> PluginRiskAdjustment:
+    item = _mapping(payload, "plugin_adjustment")
+    return PluginRiskAdjustment(
+        plugin_id=_string(_required(item, "plugin_id"), "plugin_adjustment.plugin_id"),
+        risk_multiplier=_number(_required(item, "risk_multiplier"), "plugin_adjustment.risk_multiplier"),
+        approved=_bool(item.get("approved", True), "plugin_adjustment.approved"),
+    )
+
+
+def _parse_policy(payload: object) -> AdaptiveSelectionPolicy:
+    item = _mapping(payload, "policy")
+    return AdaptiveSelectionPolicy(
+        policy_id=_string(_required(item, "policy_id"), "policy.policy_id"),
+        max_data_freshness_days=_integer(_required(item, "max_data_freshness_days"), "policy.max_data_freshness_days"),
+        minimum_regime_confidence=_number(_required(item, "minimum_regime_confidence"), "policy.minimum_regime_confidence"),
+        minimum_score=_number(_required(item, "minimum_score"), "policy.minimum_score"),
+        volatility_penalty=_number(_required(item, "volatility_penalty"), "policy.volatility_penalty"),
+        cost_penalty=_number(_required(item, "cost_penalty"), "policy.cost_penalty"),
+        max_recommendations=_integer(item.get("max_recommendations", 1), "policy.max_recommendations"),
+        fail_closed_on_unknown_regime=_bool(item.get("fail_closed_on_unknown_regime", True), "policy.fail_closed_on_unknown_regime"),
+    )
+
+
+def build_shadow_selection(payload: Mapping[str, object]) -> SelectionDecision:
+    """Validate one versioned input bundle and produce a no-order decision record."""
+    if payload.get("schema") != SELECTION_INPUT_SCHEMA:
+        raise ValueError(f"schema must equal {SELECTION_INPUT_SCHEMA}")
+    return select_shadow(
+        decision_id=_string(_required(payload, "decision_id"), "decision_id"),
+        created_at=_datetime(_required(payload, "created_at"), "created_at"),
+        market_context=_parse_market_context(_required(payload, "market_context")),
+        candidates=[_parse_candidate(item) for item in _sequence(_required(payload, "candidates"), "candidates")],
+        platform_health=[
+            _parse_platform_health(item)
+            for item in _sequence(_required(payload, "platform_health"), "platform_health")
+        ],
+        plugin_adjustments=[
+            _parse_plugin_adjustment(item)
+            for item in _sequence(payload.get("plugin_adjustments", []), "plugin_adjustments")
+        ],
+        policy=_parse_policy(_required(payload, "policy")),
+    )
+
+
+def load_shadow_selection_input(path: str | Path) -> Mapping[str, object]:
+    """Read a JSON input bundle without accepting executable configuration."""
+    try:
+        payload = json.loads(Path(path).read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise ValueError(f"cannot read selection input: {path}") from exc
+    return _mapping(payload, "selection input")
+
+
+__all__ = ["SELECTION_INPUT_SCHEMA", "build_shadow_selection", "load_shadow_selection_input"]
diff --git a/tests/test_adaptive_allocation_io.py b/tests/test_adaptive_allocation_io.py
new file mode 100644
index 0000000..501e95e
--- /dev/null
+++ b/tests/test_adaptive_allocation_io.py
@@ -0,0 +1,108 @@
+import json
+
+import pytest
+
+from quant_platform_kit.adaptive_allocation import (
+    SELECTION_INPUT_SCHEMA,
+    build_shadow_selection,
+    load_shadow_selection_input,
+)
+from quant_platform_kit.adaptive_allocation.cli import main
+
+
+def _payload(**overrides):
+    values = {
+        "schema": SELECTION_INPUT_SCHEMA,
+        "decision_id": "shadow-us-equity-001",
+        "created_at": "2026-08-29T00:00:00Z",
+        "market_context": {
+            "schema": "qsl.market_context_snapshot.v1",
+            "as_of": "2026-08-28",
+            "domain": "us_equity",
+            "data_version": "trusted-snapshot-sha",
+            "data_freshness_days": 0,
+            "regime": "normal",
+            "regime_confidence": 0.9,
+            "factors": {"momentum": 0.1},
+        },
+        "candidates": [
+            {
+                "strategy_profile": "candidate_a",
+                "release_digest": "sha256:candidate-a",
+                "lifecycle_stage": "shadow_active",
+                "approved_for_shadow": True,
+                "base_score": 0.4,
+                "estimated_volatility": 0.2,
+                "factor_exposures": {"momentum": 0.5},
+                "required_plugins": ["market_regime_control"],
+                "allowed_platform_ids": ["paper_platform"],
+            }
+        ],
+        "platform_health": [
+            {
+                "schema": "qsl.platform_health_snapshot.v1",
+                "platform_id": "paper_platform",
+                "observed_at": "2026-08-29T00:00:00+00:00",
+                "healthy": True,
+                "shadow_capable": True,
+                "reconciliation_ok": True,
+                "capacity_score": 0.8,
+                "expected_cost_bps": 1.0,
+            }
+        ],
+        "plugin_adjustments": [
+            {"plugin_id": "market_regime_control", "risk_multiplier": 0.8, "approved": True}
+        ],
+        "policy": {
+            "policy_id": "shadow-policy-v1",
+            "max_data_freshness_days": 1,
+            "minimum_regime_confidence": 0.6,
+            "minimum_score": 0.1,
+            "volatility_penalty": 0.5,
+            "cost_penalty": 0.01,
+        },
+    }
+    return values | overrides
+
+
+def test_build_shadow_selection_validates_a_versioned_bundle_and_keeps_no_order():
+    decision = build_shadow_selection(_payload())
+
+    result = decision.to_dict()
+    assert result["authority"] == "shadow_only"
+    assert result["no_order"] is True
+    assert result["recommended_strategy_profile"] == "candidate_a"
+    assert result["candidates"][0]["proposed_weight"] == 0.0
+
+
+def test_build_shadow_selection_rejects_naive_platform_timestamp():
+    payload = _payload()
+    payload["platform_health"][0]["observed_at"] = "2026-08-29T00:00:00"
+
+    with pytest.raises(ValueError, match="timezone"):
+        build_shadow_selection(payload)
+
+
+def test_build_shadow_selection_requires_versioned_context_and_integer_freshness():
+    missing_schema = _payload()
+    del missing_schema["market_context"]["schema"]
+    fractional_freshness = _payload()
+    fractional_freshness["market_context"]["data_freshness_days"] = 0.5
+
+    with pytest.raises(ValueError, match="schema"):
+        build_shadow_selection(missing_schema)
+    with pytest.raises(ValueError, match="integer"):
+        build_shadow_selection(fractional_freshness)
+
+
+def test_cli_writes_only_a_json_decision_artifact(tmp_path):
+    source = tmp_path / "input.json"
+    output = tmp_path / "output" / "decision.json"
+    source.write_text(json.dumps(_payload()), encoding="utf-8")
+
+    assert main(["--input", str(source), "--output", str(output)]) == 0
+
+    result = json.loads(output.read_text(encoding="utf-8"))
+    assert result["no_order"] is True
+    assert result["candidates"][0]["proposed_weight"] == 0.0
+    assert load_shadow_selection_input(source)["schema"] == SELECTION_INPUT_SCHEMA