diff --git a/README.md b/README.md
index b5cb60b..e130031 100644
--- a/README.md
+++ b/README.md
@@ -41,6 +41,7 @@ Production schedules should live in domain repositories and call this CLI or the
 quant-lifecycle monitor --domain us_equity
 quant-lifecycle drift --domain us_equity
 quant-lifecycle autopilot --domain us_equity --dry-run
+quant-lifecycle evidence --file path/to/evidence.json
 quant-lifecycle dashboard --format all
 ```
 
@@ -94,6 +95,9 @@ rw.update_secret("my-token", "new-token-value")
 ```
 
 ## Useful docs
+- [`docs/strategy_lifecycle_policy.md`](docs/strategy_lifecycle_policy.md)
+- [`docs/strategy_portfolio_action_matrix.md`](docs/strategy_portfolio_action_matrix.md)
+- [`docs/evidence_package_template.md`](docs/evidence_package_template.md)
 
 - [`docs/platform_notification_outcomes.md`](docs/platform_notification_outcomes.md)
 - [`docs/platform_notification_outcomes.zh-CN.md`](docs/platform_notification_outcomes.zh-CN.md)
diff --git a/README.zh-CN.md b/README.zh-CN.md
index 7e38848..004547d 100644
--- a/README.zh-CN.md
+++ b/README.zh-CN.md
@@ -41,9 +41,16 @@ python -m pytest -q
 quant-lifecycle monitor --domain us_equity
 quant-lifecycle drift --domain us_equity
 quant-lifecycle autopilot --domain us_equity --dry-run
+quant-lifecycle evidence --file path/to/evidence.json
 quant-lifecycle dashboard --format all
 ```
 
+## 延伸文档
+
+- [`docs/strategy_lifecycle_policy.zh-CN.md`](docs/strategy_lifecycle_policy.zh-CN.md)
+- [`docs/strategy_portfolio_action_matrix.zh-CN.md`](docs/strategy_portfolio_action_matrix.zh-CN.md)
+- [`docs/evidence_package_template.zh-CN.md`](docs/evidence_package_template.zh-CN.md)
+
 ## 云服务抽象层
 
 `quant_platform_kit.cloud` 包为常用云服务定义了协议接口——密钥管理、对象存储、文档数据库、计算发现和部署上下文。平台代码可以通过这些接口编写，无需硬编码到特定云厂商。
diff --git a/docs/ARCHITECTURE.md b/docs/ARCHITECTURE.md
index 2584419..4b0852a 100644
--- a/docs/ARCHITECTURE.md
+++ b/docs/ARCHITECTURE.md
@@ -145,7 +145,7 @@ Strategy profiles expose a common contract:
 | `name` | str | Profile name (snake_case) |
 | `domain` | str | Market domain (`us_equity`, `hk_equity`, `crypto`, `quant_combo`) |
 | `compatible_platforms` | list[str] | Allowed platform IDs |
-| `status` | str | `research_enabled` or `runtime_enabled` |
+| `status` | str | lifecycle stage, for example `research_backtest_only`, `shadow_candidate`, `live_candidate`, or `runtime_enabled` |
 | `evaluate()` | function | The pure decision function: `(ctx: StrategyContext) -> StrategyDecision` |
 
 Strategy packages and their coverage:
@@ -158,7 +158,7 @@ Strategy packages and their coverage:
 | `crypto-strategies` | Crypto | Live pool rotation, BTC DCA, trend rotation, equity combo | BinancePlatform |
 | `quant-us-combo-strategies` | US Combo | Russell Top50 + IBIT (50/50), leveraged combo (TQQQ/SOXL/BOXX + SPY MA200) | Schwab, IBKR, LongBridge, Firstrade |
 
-Catalog metadata is published through `get_runtime_enabled_profiles()` which platforms call to discover available strategies. Status promotion gates (`research_enabled` -> `runtime_enabled`) require passing evidence checks (backtest performance, drawdown limits, slippage tolerance).
+Catalog metadata is published through lifecycle-aware profile helpers which platforms call to discover available strategies. Status promotion gates (`research_backtest_only` -> `shadow_candidate` -> `live_candidate` -> `runtime_enabled`) require passing evidence checks (backtest performance, drawdown limits, slippage tolerance).
 
 ### 2.3 Snapshot Layer
 
@@ -375,7 +375,7 @@ Exclusion lists handle strategies that are technically compatible but not yet ap
 
 ```python
 SCHWAB_EXCLUDED_LIVE_PROFILES = [
-    "tecl_xlk_trend_income",          # research_enabled, not cleared for live
+    "tecl_xlk_trend_income",          # research_backtest_only, not cleared for live
     "soxl_soxx_trend_income",         # awaiting promo gate signoff
 ]
 ```
diff --git a/docs/STRATEGY_PLATFORM_SPEC.md b/docs/STRATEGY_PLATFORM_SPEC.md
index ec8651c..317d450 100644
--- a/docs/STRATEGY_PLATFORM_SPEC.md
+++ b/docs/STRATEGY_PLATFORM_SPEC.md
@@ -29,6 +29,16 @@ auto-sync layer described below.
 
 ## Adding a New Strategy
 
+See also: [`strategy_lifecycle_policy.md`](./strategy_lifecycle_policy.md) for the
+promotion ladder and live-enable gate model.
+
+Before requesting live enablement, attach an evidence package that covers:
+
+- backtest summary
+- drift / regime notes
+- platform compatibility evidence
+- plugin gate status, if applicable
+
 1. Add strategy definition to the shared catalog (`us_equity_strategies` or `hk_equity_strategies`)
 2. Set `compatible_platforms` to list which platforms support it
 3. Add to `runtime_enabled_profiles` for the rollout
diff --git a/docs/evidence_package_template.md b/docs/evidence_package_template.md
new file mode 100644
index 0000000..a81f6b5
--- /dev/null
+++ b/docs/evidence_package_template.md
@@ -0,0 +1,43 @@
+# Strategy Promotion Evidence Package
+
+This template defines the minimum package to request `live_candidate` review.
+
+## Required contents
+
+- strategy profile name
+- target platform(s)
+- backtest summary
+- drift / regime notes
+- platform compatibility evidence
+- plugin gate status, if applicable
+- operator notes and rollout constraints
+
+## Suggested structure
+
+```text
+profile: cn_chinext_growth_momentum_quality
+market: cn_equity
+requested_stage: live_candidate
+
+1. Backtest summary
+2. Drift and regime observations
+3. Risk review
+4. Platform compatibility evidence
+5. Plugin gate status
+6. Rollout notes
+```
+
+## Acceptance rule
+
+If any of the following are missing, keep the profile out of live settings:
+
+- the profile is not runtime-compatible with the target platform
+- the plugin gate is not clearly approved or notification-only
+- the evidence does not cover both performance and regime sensitivity
+- the request is based on a single good window only
+
+## Ownership
+
+- strategy repo: produces the evidence package
+- platform repo: verifies runtime compatibility and gate status
+- operator review: makes the final live decision
diff --git a/docs/evidence_package_template.zh-CN.md b/docs/evidence_package_template.zh-CN.md
new file mode 100644
index 0000000..8792610
--- /dev/null
+++ b/docs/evidence_package_template.zh-CN.md
@@ -0,0 +1,43 @@
+# 策略晋级证据包
+
+这份模板定义了请求 `live_candidate` 复核时最少应提交的内容。
+
+## 必要内容
+
+- 策略 profile 名称
+- 目标平台
+- 回测摘要
+- drift / regime 观察
+- 平台兼容性证据
+- 插件门槛状态（如适用）
+- 操作备注与 rollout 限制
+
+## 建议结构
+
+```text
+profile: cn_chinext_growth_momentum_quality
+market: cn_equity
+requested_stage: live_candidate
+
+1. 回测摘要
+2. drift 与 regime 观察
+3. 风险复核
+4. 平台兼容性证据
+5. 插件门槛状态
+6. rollout 备注
+```
+
+## 接受规则
+
+如果缺少以下任一项，就应继续留在 live 之外：
+
+- profile 与目标平台不兼容
+- 插件门槛不是明确 approved 或 notification-only
+- 证据没有同时覆盖收益表现和 regime 敏感性
+- 只依赖单一好窗口
+
+## 责任划分
+
+- 策略仓库：产出证据包
+- 平台仓库：验证 runtime 兼容性和门槛状态
+- 操作审批：做最终 live 决策
diff --git a/docs/strategy_lifecycle_policy.md b/docs/strategy_lifecycle_policy.md
new file mode 100644
index 0000000..aaa917a
--- /dev/null
+++ b/docs/strategy_lifecycle_policy.md
@@ -0,0 +1,84 @@
+# Strategy Lifecycle Policy
+
+[简体中文](./strategy_lifecycle_policy.zh-CN.md)
+
+This document defines the lifecycle gates for strategy profiles across the
+quant repositories.
+
+## Design Goal
+
+The lifecycle should be permissive for research and monitoring, but strict for
+capital impact.
+
+- AI monitoring may accelerate review and highlight drift.
+- AI monitoring must not bypass the live-enable gate.
+- A strategy can be observed earlier than it can trade.
+- Live enablement remains a platform decision, not just a backtest decision.
+
+## Recommended Lifecycle Stages
+
+| Stage | Meaning | Capital impact | Typical owner |
+| --- | --- | --- | --- |
+| `research_backtest_only` | Backtests, feature work, or evidence collection only | none | strategy repo |
+| `ai_monitored_candidate` | Eligible for AI review, drift scoring, and shadow tracking | none | strategy lifecycle |
+| `shadow_candidate` | Shadow runs are stable enough for repeatable comparisons | none | strategy lifecycle |
+| `live_candidate` | Passed validation and is awaiting platform enablement | gated | platform + strategy |
+| `runtime_enabled` | Exposed by `get_runtime_enabled_profiles()` and allowed in runtime settings | yes | platform repo |
+
+### Practical interpretation
+
+- `research_backtest_only` is the default for anything new.
+- `ai_monitored_candidate` is the lowest-friction review stage when the
+  organization already has automated monitoring.
+- `shadow_candidate` should require repeatable shadow consistency, not just one
+  good backtest.
+- `live_candidate` should only be used when the strategy has enough evidence to
+  justify platform enablement.
+- `runtime_enabled` is the only stage that should influence live deployment
+  defaults.
+
+## Three-Gate Rule
+
+A strategy should clear all three gates before live use:
+
+1. **Strategy gate**
+   - Does the strategy have enough history, risk characterization, and drift
+     tolerance to move beyond research?
+2. **Plugin gate**
+   - If the strategy depends on plugins, are those plugins at least
+     `automation_approved` or explicitly `notification_only`?
+3. **Platform gate**
+   - Does the target platform expose the profile via
+     `get_runtime_enabled_profiles()` and accept the required runtime inputs?
+
+Any one of these gates failing should keep the profile out of live settings.
+
+## Recommended Promotion Policy
+
+- Keep the monitoring threshold relatively low so candidates are visible early.
+- If AI monitoring already exists, use it to move promising profiles into
+  `ai_monitored_candidate` quickly; this stage is for visibility, not capital.
+- Keep the live-enable threshold high so runtime exposure remains deliberate.
+- Prefer promotion by evidence package, not by ad hoc overrides. A promotion
+  package should include backtest summary, drift notes, risk review, and
+  platform compatibility evidence.
+- See [`evidence_package_template.md`](./evidence_package_template.md) for the
+  recommended package shape.
+- When a strategy is a wrapper or orchestrator, promote it only after the
+  wrapped components are stable and the wrapper itself has been validated.
+
+## Repo-Level Guidance
+
+- **US equity**: long-history trend / rotation profiles can move through the
+  lifecycle earlier; wrapper combos should stay candidate-first.
+- **HK equity**: keep live exposure narrow and promote only stable runtime
+  profiles.
+- **CN equity**: treat QMT-specific optional runtime profiles separately from
+  the main live catalog.
+- **Crypto**: keep the monitoring stage permissive, but use a stricter live
+  gate because regime shifts are faster.
+
+## Operational Rule
+
+If a profile is not returned by `get_runtime_enabled_profiles()`, it should
+stay out of live runtime settings regardless of monitoring status.
diff --git a/docs/strategy_lifecycle_policy.zh-CN.md b/docs/strategy_lifecycle_policy.zh-CN.md
new file mode 100644
index 0000000..043f7ef
--- /dev/null
+++ b/docs/strategy_lifecycle_policy.zh-CN.md
@@ -0,0 +1,70 @@
+# 策略生命周期门槛策略
+
+[English](./strategy_lifecycle_policy.md)
+
+本文定义量化仓库中策略 profile 的生命周期门槛。
+
+## 设计目标
+
+生命周期应当对研究和监控保持宽松，但对资金影响保持严格。
+
+- AI 监控可以加快复核和暴露 drift。
+- AI 监控不能绕过 live 启用门槛。
+- 策略可以先被观察，再允许交易。
+- live 启用仍然是平台决策，不只是回测结论。
+
+## 推荐生命周期阶段
+
+| 阶段 | 含义 | 资金影响 | 常见归属 |
+| --- | --- | --- | --- |
+| `research_backtest_only` | 仅回测、特征开发或证据收集 | 无 | 策略仓库 |
+| `ai_monitored_candidate` | 可进入 AI 复核、drift 打分和 shadow 跟踪 | 无 | 策略生命周期 |
+| `shadow_candidate` | shadow 运行已经稳定，可做重复对比 | 无 | 策略生命周期 |
+| `live_candidate` | 已通过验证，等待平台启用 | 受门槛控制 | 平台 + 策略 |
+| `runtime_enabled` | 会被 `get_runtime_enabled_profiles()` 暴露并允许进入运行时设置 | 有 | 平台仓库 |
+
+### 实际含义
+
+- `research_backtest_only` 适合作为所有新策略的默认起点。
+- 如果组织已经有自动监控，`ai_monitored_candidate` 可以作为最低摩擦的复核阶段。
+- `shadow_candidate` 应该要求重复 shadow 一致性，而不是只看一次漂亮回测。
+- `live_candidate` 只适合证据足够支持平台启用的策略。
+- `runtime_enabled` 才应该影响 live 配置默认值。
+
+## 三道门槛
+
+策略进入 live 之前，应同时通过三道门槛：
+
+1. **策略门槛**
+   - 是否具备足够的历史、风险画像和 drift 容忍度，能从研究阶段前进？
+2. **插件门槛**
+   - 如果策略依赖插件，这些插件是否至少是 `automation_approved`，
+     或者被明确标记为 `notification_only`？
+3. **平台门槛**
+   - 目标平台是否通过 `get_runtime_enabled_profiles()` 暴露该 profile，
+     并接受所需运行时输入？
+
+任意一项不过关，都应该继续留在 live 之外。
+
+## 推荐晋级策略
+
+- 监控阈值可以相对低一些，让候选策略尽早可见。
+- 如果组织已经有自动 AI 监控，就用它尽快把合适的 profile 推到
+  `ai_monitored_candidate`；这一层只为可见性，不为资金。
+- live 启用阈值要保持高一些，确保运行时暴露是明确决策。
+- 以证据包晋级，而不是临时 override。证据包至少应包含回测摘要、
+  drift 记录、风险复核和平台兼容性证据。
+- 证据包建议结构见 [`evidence_package_template.zh-CN.md`](./evidence_package_template.zh-CN.md)。
+- 如果策略是 wrapper / orchestrator，应先确认被包装组件稳定，再考虑给 wrapper 晋级。
+
+## 仓库级建议
+
+- **美股**：长历史趋势 / 轮动策略可以更早进入生命周期推进；wrapper combo 应优先保持 candidate 状态。
+- **港股**：保持 live 暴露范围窄，只晋级稳定的 runtime profile。
+- **A 股**：把 QMT 专用的可选 runtime profile 和主 live catalog 分开看待。
+- **加密货币**：监控阶段可以宽松一些，但 live 门槛应更严格，因为 regime 切换更快。
+
+## 运行规则
+
+如果某个 profile 没有被 `get_runtime_enabled_profiles()` 返回，那么它就不应该进入 live runtime settings，
+无论监控状态是否已经打开。
diff --git a/docs/strategy_portfolio_action_matrix.md b/docs/strategy_portfolio_action_matrix.md
new file mode 100644
index 0000000..e1526ba
--- /dev/null
+++ b/docs/strategy_portfolio_action_matrix.md
@@ -0,0 +1,103 @@
+# Strategy Portfolio Action Matrix
+
+[简体中文](./strategy_portfolio_action_matrix.zh-CN.md)
+
+This document turns the current cross-repository review into an execution plan.
+It uses the present repo state as the source of truth and separates:
+
+- keep tuning
+- rebuild / redesign
+- deprecate / downgrade
+- new strategy ideas
+
+## Decision rules
+
+- If a profile is still `runtime_enabled` and has clear evidence, keep tuning it.
+- If a profile is useful but its shape is wrong, keep the idea and rebuild the
+  wrapper / orchestrator.
+- If a profile is clearly weaker than the main live line, downgrade it to
+  `shadow_candidate`, `research_backtest_only`, or archive it.
+- If a market lacks a good live candidate, prefer a new architecture instead of
+  adding more variants.
+
+## Cross-market matrix
+
+| Market | Keep tuning | Rebuild / redesign | Downgrade / retire | New ideas worth testing |
+| --- | --- | --- | --- | --- |
+| US equity | `global_etf_rotation`, `tqqq_growth_income`, `soxl_soxx_trend_income`, `russell_top50_leader_rotation`, DCA lines | `us_equity_combo`, `us_equity_combo_leveraged` should stay as shadow/orchestrator layers rather than live profiles | `tecl_xlk_trend_income` stays research-only | Optional LEAPS overlay, but only as a separately gated research track |
+| HK equity | `hk_global_etf_tactical_rotation`, `hk_low_vol_dividend_quality_snapshot` | `hk_equity_combo` should be a research/orchestration wrapper, not a live profile | any combo-style live promotion should be blocked until evidence improves | A dedicated factor-enhanced rotation wrapper only after the core lines are stable |
+| CN equity | `cn_industry_etf_rotation`, `cn_industry_etf_rotation_aggressive` | `cn_equity_combo` should be rebuilt as an orchestrator / dual-track composition, not a direct live profile | `cn_index_etf_tactical_rotation` stays legacy/research; `cn_chinext_tactical_rotation` and `cn_chinext_growth_momentum_quality_snapshot` should be redesigned, not discarded | `cn_dual_track_combo`, a first-class ChiNext growth sleeve, and a separate STAR growth sleeve |
+| Crypto | `crypto_live_pool_rotation`, `crypto_btc_dca` | `crypto_trend_rotation`, `crypto_equity_combo` need redesign before any live promotion | all non-live crypto profiles stay research/shadow until proven otherwise | A volatility-regime filter or stablecoin cash-deploy layer before adding more rotation variants |
+
+## What this means by market
+
+### US
+
+US already has a strong main live set. The right move is not more live variants;
+it is better separation:
+
+- keep the live trend and rotation lines as the core
+- keep combo logic as a shadow/orchestrator layer
+- retire weak legacy templates like TECL
+
+### HK
+
+HK should stay narrow:
+
+- keep the ETF rotation and snapshot-backed dividend-quality line
+- treat combo as a research wrapper
+- do not broaden live exposure until the evidence gap is closed
+
+### CN
+
+CN still has room for improvement, but the current strongest path is:
+
+- one direct ETF rotation main line
+- one aggressive variant under controlled rollout
+- one future orchestrator / dual-track line
+
+ChiNext and STAR should not be treated as disposable research branches; they
+are board-specific growth sleeves whose design needs to match the market
+regime.
+
+The combo profile should not compete with the main live line until it is
+rebuilt as a composition engine.
+
+### Crypto
+
+Crypto is the most regime-sensitive domain, so the live gate should stay strict:
+
+- keep the live pool rotation line as the main runtime strategy
+- keep BTC DCA as a simple tuneable accumulation line
+- redesign trend and combo variants before any live promotion
+
+## Plugin and gate implications
+
+- `notification_only` is fine for monitoring and research visibility.
+- `automation_candidate` is fine for shadow and pre-live validation.
+- `automation_approved + position_control_allowed` remains the hard gate for
+  any automatic position impact.
+- A strategy being AI-monitored is not the same as being live-enabled.
+
+## Recommended next implementation steps
+
+1. Keep the current live lines stable.
+2. Rebuild combo/orchestrator profiles around composition, not duplication.
+3. Add any new strategies only if they solve a real gap in the matrix above.
+4. Use evidence packages to promote, not ad hoc exceptions.
+
+## Reference strategy families
+
+The external research that best matches this matrix is still:
+
+- time-series momentum / trend following
+- factor investing: quality, momentum, low volatility, dividend yield
+- multi-asset value + momentum overlays
+
+Official references:
+
+- AQR Trends Everywhere
+- AQR Value and Momentum Everywhere
+- Kenneth French Data Library
+- MSCI Factor Indexes
+- EDHEC trend-following research
diff --git a/docs/strategy_portfolio_action_matrix.zh-CN.md b/docs/strategy_portfolio_action_matrix.zh-CN.md
new file mode 100644
index 0000000..3606eed
--- /dev/null
+++ b/docs/strategy_portfolio_action_matrix.zh-CN.md
@@ -0,0 +1,97 @@
+# 策略组合行动矩阵
+
+[English](./strategy_portfolio_action_matrix.md)
+
+这份文档把当前跨仓库 review 结果整理成可执行方案。它以现有仓库状态为准，
+把结论分成：
+
+- 继续调参
+- 重构 / 重设
+- 降级 / 废弃
+- 值得新增的方向
+
+## 判定规则
+
+- 如果 profile 仍然是 `runtime_enabled`，而且证据清楚，就继续调参。
+- 如果策略思路有价值，但形态不对，就保留思路、重构 wrapper / orchestrator。
+- 如果某条线明显弱于主 live 线，就降级到 `shadow_candidate`、
+  `research_backtest_only`，或者直接归档。
+- 如果某个市场还没有合适的 live 候选，就优先改架构，不要继续堆变体。
+
+## 跨市场矩阵
+
+| 市场 | 继续调参 | 重构 / 重设 | 降级 / 废弃 | 值得测试的新方向 |
+| --- | --- | --- | --- | --- |
+| 美股 | `global_etf_rotation`、`tqqq_growth_income`、`soxl_soxx_trend_income`、`russell_top50_leader_rotation`、各 DCA 线 | `us_equity_combo`、`us_equity_combo_leveraged` 更适合作为 shadow / orchestrator 层，而不是 live profile | `tecl_xlk_trend_income` 保持 research-only | LEAPS 增强层可以继续研究，但必须单独门控 |
+| 港股 | `hk_global_etf_tactical_rotation`、`hk_low_vol_dividend_quality_snapshot` | `hk_equity_combo` 应该是 research/orchestration wrapper，不应当是 live profile | 任何 combo 型 live 推进都应先阻断，直到证据补齐 | 只有核心线稳定后，才考虑独立的因子增强 wrapper |
+| A 股 | `cn_industry_etf_rotation`、`cn_industry_etf_rotation_aggressive` | `cn_equity_combo` 应重做成 orchestrator / dual-track composition，而不是直接 live profile | `cn_index_etf_tactical_rotation` 保持 legacy / research；`cn_chinext_tactical_rotation` 和 `cn_chinext_growth_momentum_quality_snapshot` 应重设计，不是简单废弃 | `cn_dual_track_combo`、独立的创业板成长 sleeve、以及独立的科创板成长 sleeve |
+| 加密 | `crypto_live_pool_rotation`、`crypto_btc_dca` | `crypto_trend_rotation`、`crypto_equity_combo` 需要重构后再谈 live | 其余非 live crypto profile 保持 research/shadow，直到证据充分 | 先加波动率 regime 过滤或 stablecoin 资金投放层，再考虑更多轮动变体 |
+
+## 各市场解释
+
+### 美股
+
+美股已经有较强的主 live 线，重点不是再加 live 变体，而是分层更清楚：
+
+- 主 live 线继续稳定
+- combo 逻辑保留为 shadow / orchestrator
+- TECL 这类弱线继续退场
+
+### 港股
+
+港股应该保持窄而稳：
+
+- 保留 ETF 轮动和 snapshot 红利质量线
+- combo 只做 research wrapper
+- 在证据缺口补齐前，不要扩 live 面
+
+### A 股
+
+A 股还有改进空间，但当前最强路径仍然是：
+
+- 一条直接 ETF 轮动主线
+- 一条受控的 aggressive 变体
+- 一条未来可组合的 orchestrator / dual-track 线
+
+创业板和科创板不应当被当成可随手丢弃的研究分支；它们是板块级
+成长 sleeve，设计必须贴合真实市场 regime。
+
+combo 现在不该和主 live 线争位置，应该先重做成 composition engine。
+
+### 加密
+
+加密是 regime 切换最快的域，所以 live 门槛要更严格：
+
+- 保持 live pool rotation 作为主 runtime 策略
+- 保持 BTC DCA 作为简单、可调的积累线
+- trend 和 combo 变体先重构，再谈 live
+
+## 插件和门槛含义
+
+- `notification_only` 适合监控和研究可见性
+- `automation_candidate` 适合 shadow 和 pre-live 验证
+- `automation_approved + position_control_allowed` 仍然是自动仓位影响的硬门槛
+- AI monitored 不等于 live enabled
+
+## 推荐的下一步
+
+1. 先稳住当前 live 主线
+2. 把 combo / orchestrator 做成真正的 composition，而不是复制主线
+3. 新策略只在能补齐真实缺口时再加
+4. 用证据包晋级，不用临时特批
+
+## 参考策略家族
+
+和这份矩阵最吻合的外部研究主要还是：
+
+- 时间序列动量 / 趋势跟随
+- 因子投资：质量、动量、低波动、股息收益
+- 多资产 value + momentum overlay
+
+官方参考：
+
+- AQR Trends Everywhere
+- AQR Value and Momentum Everywhere
+- Kenneth French Data Library
+- MSCI Factor Indexes
+- EDHEC 趋势跟随研究
diff --git a/scripts/validate_evidence_package.py b/scripts/validate_evidence_package.py
new file mode 100644
index 0000000..fccd575
--- /dev/null
+++ b/scripts/validate_evidence_package.py
@@ -0,0 +1,12 @@
+#!/usr/bin/env python3
+"""Validate a strategy promotion evidence package."""
+
+from __future__ import annotations
+
+import sys
+
+from quant_platform_kit.strategy_lifecycle.cli import main
+
+
+if __name__ == "__main__":
+    raise SystemExit(main(["evidence", "--file", *sys.argv[1:]]) if len(sys.argv) > 1 else 2)
diff --git a/src/quant_platform_kit/strategy_lifecycle/__init__.py b/src/quant_platform_kit/strategy_lifecycle/__init__.py
index 50c09b8..c23591d 100644
--- a/src/quant_platform_kit/strategy_lifecycle/__init__.py
+++ b/src/quant_platform_kit/strategy_lifecycle/__init__.py
@@ -14,6 +14,13 @@
     UpdateStage,
     WindowPerformance,
 )
+from quant_platform_kit.strategy_lifecycle.evidence_gate import (
+    EvidenceGateResult,
+    EvidencePackage,
+    load_evidence_package,
+    validate_evidence_package,
+    validate_evidence_package_file,
+)
 
 __all__ = [
     "BacktestResult",
@@ -28,4 +35,9 @@
     "UpdateLogEntry",
     "UpdateStage",
     "WindowPerformance",
+    "EvidenceGateResult",
+    "EvidencePackage",
+    "load_evidence_package",
+    "validate_evidence_package",
+    "validate_evidence_package_file",
 ]
diff --git a/src/quant_platform_kit/strategy_lifecycle/cli.py b/src/quant_platform_kit/strategy_lifecycle/cli.py
index dfbd087..d9a1863 100644
--- a/src/quant_platform_kit/strategy_lifecycle/cli.py
+++ b/src/quant_platform_kit/strategy_lifecycle/cli.py
@@ -158,6 +158,27 @@ def _run_lifecycle(args: argparse.Namespace) -> int:
     return 0
 
 
+def _run_evidence(args: argparse.Namespace) -> int:
+    _print(f"[evidence] Validating package file={args.file}")
+    validate_evidence_package_file = _load_callable(
+        "quant_platform_kit.strategy_lifecycle.evidence_gate",
+        "validate_evidence_package_file",
+    )
+    result = validate_evidence_package_file(args.file)
+    if getattr(args, "json", False):
+        import json
+
+        _print(json.dumps(result.to_dict(), ensure_ascii=False, indent=2))
+    else:
+        status = "valid" if result.valid else "invalid"
+        _print(f"[evidence] status={status} issues={len(result.issues)} warnings={len(result.warnings)}")
+        for item in result.issues:
+            _print(f"  - issue: {item}")
+        for item in result.warnings:
+            _print(f"  - warning: {item}")
+    return 0 if result.valid else 1
+
+
 def build_parser() -> argparse.ArgumentParser:
     parser = argparse.ArgumentParser(prog="quant-lifecycle", description="Quant strategy lifecycle CLI.")
     parser.add_argument("--version", action="version", version="quant-lifecycle 0.10.0")
@@ -197,6 +218,11 @@ def build_parser() -> argparse.ArgumentParser:
     autopilot.add_argument("--no-issues", action="store_true")
     autopilot.set_defaults(func=_run_autopilot)
 
+    evidence = subparsers.add_parser("evidence", help="Validate a strategy promotion evidence package.")
+    evidence.add_argument("--file", required=True)
+    evidence.add_argument("--json", action="store_true")
+    evidence.set_defaults(func=_run_evidence)
+
     lifecycle = subparsers.add_parser("lifecycle", help="Run the full lifecycle pipeline.")
     lifecycle.add_argument("--domain", default="us_equity")
     lifecycle.add_argument("--strategy", default=None)
diff --git a/src/quant_platform_kit/strategy_lifecycle/evidence_gate.py b/src/quant_platform_kit/strategy_lifecycle/evidence_gate.py
new file mode 100644
index 0000000..37f2b07
--- /dev/null
+++ b/src/quant_platform_kit/strategy_lifecycle/evidence_gate.py
@@ -0,0 +1,237 @@
+"""Evidence package validation for lifecycle promotion requests."""
+
+from __future__ import annotations
+
+import json
+import tomllib
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Mapping
+
+ALLOWED_EVIDENCE_STAGES = (
+    "research_backtest_only",
+    "ai_monitored_candidate",
+    "shadow_candidate",
+    "live_candidate",
+    "runtime_enabled",
+)
+LIVE_EVIDENCE_STAGES = {"live_candidate", "runtime_enabled"}
+ALLOWED_PLUGIN_GATE_STATUSES = {
+    "automation_approved",
+    "deprecated_compatibility",
+    "notification_only",
+    "research_only",
+    "shadow_observer",
+}
+
+
+@dataclass(frozen=True)
+class EvidencePackage:
+    strategy_profile: str
+    domain: str
+    requested_stage: str
+    target_platforms: tuple[str, ...] = ()
+    backtest_summary: Mapping[str, Any] = field(default_factory=dict)
+    drift_notes: Any = None
+    platform_compatibility: Mapping[str, Any] | None = None
+    plugin_gate: Any = None
+    rollout_notes: Any = None
+    operator_notes: Any = None
+    evidence_version: str = "evidence_package.v1"
+    submitted_at: str = ""
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "strategy_profile": self.strategy_profile,
+            "domain": self.domain,
+            "requested_stage": self.requested_stage,
+            "target_platforms": list(self.target_platforms),
+            "backtest_summary": dict(self.backtest_summary),
+            "drift_notes": self.drift_notes,
+            "platform_compatibility": self.platform_compatibility,
+            "plugin_gate": self.plugin_gate,
+            "rollout_notes": self.rollout_notes,
+            "operator_notes": self.operator_notes,
+            "evidence_version": self.evidence_version,
+            "submitted_at": self.submitted_at,
+        }
+
+
+@dataclass(frozen=True)
+class EvidenceGateResult:
+    valid: bool
+    package: EvidencePackage
+    issues: tuple[str, ...] = ()
+    warnings: tuple[str, ...] = ()
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "valid": self.valid,
+            "issues": list(self.issues),
+            "warnings": list(self.warnings),
+            "package": self.package.to_dict(),
+        }
+
+
+def load_evidence_package(path: str | Path) -> dict[str, Any]:
+    evidence_path = Path(path)
+    raw = evidence_path.read_text(encoding="utf-8")
+    suffix = evidence_path.suffix.lower()
+    if suffix == ".json":
+        payload = json.loads(raw)
+    elif suffix == ".toml":
+        payload = tomllib.loads(raw)
+    else:
+        raise ValueError(f"Unsupported evidence package format: {evidence_path.suffix or '<none>'}")
+    if not isinstance(payload, dict):
+        raise ValueError("Evidence package must decode to a mapping")
+    return payload
+
+
+def validate_evidence_package(raw: Mapping[str, Any]) -> EvidenceGateResult:
+    issues: list[str] = []
+    warnings: list[str] = []
+
+    strategy_profile = _first_str(raw, "strategy_profile", "profile")
+    domain = _first_str(raw, "domain", "market")
+    requested_stage = _first_str(raw, "requested_stage", "stage")
+    target_platforms = _normalize_platforms(
+        _first_value(raw, "target_platforms", "platforms", "target_platform", "runtime_targets")
+    )
+    backtest_summary = _first_mapping(raw, "backtest_summary", "backtest", "evidence")
+    drift_notes = _first_value(raw, "drift_notes", "drift", "regime_notes")
+    platform_compatibility = _first_mapping(
+        raw,
+        "platform_compatibility",
+        "compatibility",
+        "platform_evidence",
+    )
+    plugin_gate = _first_value(raw, "plugin_gate", "plugin_gates", "plugins")
+    rollout_notes = _first_value(raw, "rollout_notes", "rollout", "operator_notes")
+    evidence_version = str(_first_value(raw, "evidence_version", "schema", default="evidence_package.v1") or "evidence_package.v1")
+    submitted_at = str(_first_value(raw, "submitted_at", "generated_at", "created_at", default="") or "")
+
+    if not strategy_profile:
+        issues.append("missing strategy_profile/profile")
+    if not domain:
+        issues.append("missing domain/market")
+    if requested_stage not in ALLOWED_EVIDENCE_STAGES:
+        issues.append(f"unsupported requested_stage: {requested_stage!r}")
+    if not backtest_summary:
+        issues.append("missing backtest_summary")
+    elif not _backtest_summary_has_evidence(backtest_summary):
+        issues.append("backtest_summary missing core metrics or observation count")
+
+    if requested_stage in LIVE_EVIDENCE_STAGES:
+        if not target_platforms:
+            issues.append("live request requires target_platforms/platforms")
+        if not _is_non_empty(drift_notes):
+            issues.append("live request requires drift_notes")
+        if not platform_compatibility:
+            issues.append("live request requires platform_compatibility evidence")
+        elif not _platform_compatibility_is_usable(platform_compatibility):
+            issues.append("platform_compatibility evidence is incomplete")
+        if plugin_gate is not None and not _plugin_gate_is_usable(plugin_gate):
+            issues.append("plugin_gate evidence is incomplete or unsupported")
+
+    if requested_stage in {"research_backtest_only", "ai_monitored_candidate"} and not _is_non_empty(rollout_notes):
+        warnings.append("rollout_notes is recommended for candidates")
+
+    package = EvidencePackage(
+        strategy_profile=str(strategy_profile or ""),
+        domain=str(domain or ""),
+        requested_stage=str(requested_stage or ""),
+        target_platforms=target_platforms,
+        backtest_summary=backtest_summary,
+        drift_notes=drift_notes,
+        platform_compatibility=platform_compatibility,
+        plugin_gate=plugin_gate,
+        rollout_notes=rollout_notes,
+        operator_notes=_first_value(raw, "operator_notes", "notes"),
+        evidence_version=evidence_version,
+        submitted_at=submitted_at,
+    )
+    return EvidenceGateResult(valid=not issues, package=package, issues=tuple(issues), warnings=tuple(warnings))
+
+
+def validate_evidence_package_file(path: str | Path) -> EvidenceGateResult:
+    return validate_evidence_package(load_evidence_package(path))
+
+
+def _first_value(raw: Mapping[str, Any], *keys: str, default: Any = None) -> Any:
+    for key in keys:
+        if key in raw and raw[key] is not None:
+            return raw[key]
+    return default
+
+
+def _first_mapping(raw: Mapping[str, Any], *keys: str) -> Mapping[str, Any]:
+    value = _first_value(raw, *keys)
+    return value if isinstance(value, Mapping) else {}
+
+
+def _first_str(raw: Mapping[str, Any], *keys: str) -> str:
+    value = _first_value(raw, *keys)
+    return str(value).strip() if value is not None else ""
+
+
+def _normalize_platforms(value: Any) -> tuple[str, ...]:
+    if value is None:
+        return ()
+    if isinstance(value, str):
+        items = [value]
+    elif isinstance(value, (list, tuple, set, frozenset)):
+        items = list(value)
+    else:
+        items = [value]
+    return tuple(str(item).strip() for item in items if str(item).strip())
+
+
+def _is_non_empty(value: Any) -> bool:
+    if value is None:
+        return False
+    if isinstance(value, str):
+        return bool(value.strip())
+    if isinstance(value, Mapping):
+        return bool(value)
+    if isinstance(value, (list, tuple, set, frozenset)):
+        return bool(value)
+    return True
+
+
+def _backtest_summary_has_evidence(summary: Mapping[str, Any]) -> bool:
+    if not summary:
+        return False
+    if summary.get("observation_count") not in (None, "", 0):
+        return True
+    if _is_non_empty(summary.get("windows")):
+        return True
+    for key in ("cagr", "sharpe_ratio", "total_return", "max_drawdown"):
+        if summary.get(key) not in (None, ""):
+            return True
+    return False
+
+
+def _platform_compatibility_is_usable(compatibility: Mapping[str, Any]) -> bool:
+    if not compatibility:
+        return False
+    if any(bool(compatibility.get(key)) for key in ("verified", "compatible", "supported", "enabled")):
+        return True
+    if _is_non_empty(compatibility.get("runtime_enabled_profiles")):
+        return True
+    if _is_non_empty(compatibility.get("target_platforms")):
+        return True
+    return False
+
+
+def _plugin_gate_is_usable(plugin_gate: Any) -> bool:
+    if isinstance(plugin_gate, Mapping):
+        status = str(plugin_gate.get("status") or plugin_gate.get("evidence_status") or "").strip().lower()
+        if status and status not in ALLOWED_PLUGIN_GATE_STATUSES:
+            return False
+        if status == "automation_approved":
+            return bool(plugin_gate.get("position_control_allowed", False))
+        return True
+    if isinstance(plugin_gate, (list, tuple)):
+        return all(_plugin_gate_is_usable(item) for item in plugin_gate)
+    return False
diff --git a/tests/test_lifecycle_evidence_gate.py b/tests/test_lifecycle_evidence_gate.py
new file mode 100644
index 0000000..04c25a5
--- /dev/null
+++ b/tests/test_lifecycle_evidence_gate.py
@@ -0,0 +1,80 @@
+"""Tests for strategy_lifecycle.evidence_gate."""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+from quant_platform_kit.strategy_lifecycle.cli import main
+from quant_platform_kit.strategy_lifecycle.evidence_gate import (
+    load_evidence_package,
+    validate_evidence_package,
+    validate_evidence_package_file,
+)
+
+
+def test_research_package_valid() -> None:
+    result = validate_evidence_package(
+        {
+            "profile": "cn_equity_combo",
+            "market": "cn_equity",
+            "requested_stage": "research_backtest_only",
+            "backtest_summary": {"observation_count": 252, "sharpe_ratio": 1.2},
+        }
+    )
+
+    assert result.valid
+    assert result.package.strategy_profile == "cn_equity_combo"
+
+
+def test_live_package_requires_compatibility_and_drift() -> None:
+    result = validate_evidence_package(
+        {
+            "strategy_profile": "cn_chinext_growth_momentum_quality",
+            "domain": "cn_equity",
+            "requested_stage": "live_candidate",
+            "target_platforms": ["qmt"],
+            "backtest_summary": {"observation_count": 252, "sharpe_ratio": 1.1},
+        }
+    )
+
+    assert not result.valid
+    assert "drift_notes" in " ".join(result.issues)
+    assert "platform_compatibility" in " ".join(result.issues)
+
+
+def test_live_package_valid_when_required_sections_present() -> None:
+    result = validate_evidence_package(
+        {
+            "strategy_profile": "cn_chinext_growth_momentum_quality",
+            "domain": "cn_equity",
+            "requested_stage": "live_candidate",
+            "target_platforms": ["qmt"],
+            "backtest_summary": {"observation_count": 252, "sharpe_ratio": 1.1},
+            "drift_notes": {"status": "watch", "summary": "stable"},
+            "platform_compatibility": {"verified": True},
+            "plugin_gate": {"status": "notification_only"},
+        }
+    )
+
+    assert result.valid
+
+
+def test_file_loader_and_cli(tmp_path: Path) -> None:
+    payload = {
+        "strategy_profile": "cn_equity_combo",
+        "domain": "cn_equity",
+        "requested_stage": "research_backtest_only",
+        "backtest_summary": {"observation_count": 126, "total_return": 0.12},
+    }
+    path = tmp_path / "evidence.json"
+    path.write_text(json.dumps(payload), encoding="utf-8")
+
+    loaded = load_evidence_package(path)
+    assert loaded["strategy_profile"] == "cn_equity_combo"
+
+    result = validate_evidence_package_file(path)
+    assert result.valid
+
+    exit_code = main(["evidence", "--file", str(path)])
+    assert exit_code == 0