@mstar-harness/opencode 0.7.9 → 1.0.1

package/harness-commands/iteration-drive.md

@@ -30,6 +30,7 @@ Phase 2: Autonomous Execute  →  Phase 3: iteration-close  →  Phase 4: Create
 | 禁止（PM 线程） | 必须 |
 |-----------------|------|
 | Write/Edit/Shell 产品代码、写测试、跑 QC 审查（Phase 2） | 每条 implement/QC/QA Assignment ⇒ **1 次 `Task`** |
+| **多 task plan 用 inline 大包派发**（整份 plan / T1–Tn 贴进一个 dev Assignment） | **SDD**：`mstar-sdd` per-task 循环 — `task-brief` → implementer → `review-package` → task reviewer → `progress.md` |
 | 只写 Assignment 就进入下一 gate | 同轮 dispatch：`Subagent invokes issued: N`（N = Assignment 条数） |
 | 最后一个 plan `Done` 后直接开 PR / 汇报结束 | **Phase 3 → 4 → 5** 顺序执行 |
 | Phase 4 开 PR 后停止 | Phase 5 loop 至 merge-ready；**禁止**未过 §5.5 就结束会话 |
@@ -50,11 +51,12 @@ Phase 2: Autonomous Execute  →  Phase 3: iteration-close  →  Phase 4: Create
 
 1. `mstar-harness-core`
 2. `mstar-roles` → `references/project-manager.md`
-3. `skills/pm/SKILL.md` → **§ Host entry** + **§ Boot**（PM role identity + dispatch-first rules）
-4. `mstar-iteration` → **§ Phase 2–5**（§3 close gate；§4 PR delivery；§5 merge-ready loop）
-5. `mstar-compound` — Phase 3 §3.2 前加载（含 Phase 6 索引登记）
-6. `mstar-dispatch-gates` + host reference
-7. `mstar-plan-artifacts`, `mstar-plan-conventions`, `mstar-branch-worktree`
+3. `mstar-iteration` → **§ Phase 2–5**（§3 close gate；§4 PR delivery；§5 merge-ready loop）
+4. `mstar-compound` — Phase 3 §3.2 前加载（含 Phase 6 索引登记）
+5. `mstar-dispatch-gates` + host reference
+6. **`mstar-sdd`** — **before first implement dispatch** in Phase 2（per-task loop SSOT；多 task plan 默认 `Execution mode: sdd`）
+7. `mstar-review-qc` — before first QC dispatch in Phase 2
+8. `mstar-plan-artifacts`, `mstar-plan-conventions`, `mstar-branch-worktree`
 
 ## Phase 2: Autonomous Execute
 
@@ -74,9 +76,17 @@ Execute **`mstar-iteration` § Phase 2** exactly. Summary:
 4. **Integration branch** (§ 2.3) — `git checkout -b <spec_integration_branch> <iteration_base_branch>` when creating（**not** implicit `main`）
 5. **Per-plan loop** (§ 2.4) — for each non-`Done` plan:
    - Create plan feature branch from integration
-   - Dispatch implement subagents (dispatch-first)
+   - **SDD implement**（默认 `Execution mode: sdd`；`mstar-sdd` 已载入）— per task **串行**：
+     1. `sdd-workspace <plan-id>` → `{SDD_DIR}`
+     2. `task-brief <plan> N` → `{SDD_DIR}/task-N-brief.md`
+     3. Dispatch **one** implementer subagent（brief + report 路径；`references/implementer-prompt.md` 或 sticky 时 `implementer-continuation-prompt.md`）
+     4. `review-package BASE HEAD` → task diff
+     5. Dispatch **one** task reviewer subagent
+     6. Append `{SDD_DIR}/progress.md`；更新 `status.json` `task_commits[]`（如使用）
+     7. Next task — **never** parallel implementers；同 plan 同 dev 可用 **`SDD implementer session: sticky`**（`mstar-sdd/references/sticky-implementer-session.md`）
+   - **禁止**：把整份 plan 或 T1–Tn 全文贴进 implement dispatch；文件交接 SSOT → `mstar-sdd/references/file-handoffs.md`
    - Update `status.json` + main plan after each Completion Report v2
-   - QC tri-review + QA per `mstar-review-qc`
+   - QC **full tri-review** (`QC mode: full tri-review`, **N=3**, `mstar-review-qc`) + branch `review-package` + QA per plan
    - Merge plan branch → integration branch
    - Cross-plan progress sync → compass
    - Next plan

package/harness-commands/iteration-start.md

@@ -28,12 +28,11 @@ Detailed workflow → **`mstar-iteration` § Phase 1: iteration-start**（含 §
 
 1. `mstar-harness-core`
 2. `mstar-roles` → `references/project-manager.md`
-3. `skills/pm/SKILL.md` → **§ Host entry** + **§ Boot**
-4. `mstar-iteration` → **§ Phase 1: iteration-start**（迭代范围、compass 模板、§1.6 Review & Edit chain、状态初始化）
-5. `mstar-dispatch-gates`
-6. `mstar-phase-gates` → Prepare（specify → clarify → plan）
-7. `mstar-plan-conventions`, `mstar-plan-artifacts`
-8. `mstar-host` → active host reference（invoke 能力）
+3. `mstar-iteration` → **§ Phase 1: iteration-start**（迭代范围、compass 模板、§1.6 Review & Edit chain、状态初始化）
+4. `mstar-dispatch-gates`
+5. `mstar-phase-gates` → Prepare（specify → clarify → plan）
+6. `mstar-plan-conventions`, `mstar-plan-artifacts`
+7. `mstar-host` → active host reference（invoke 能力）
 
 ## 1. Research
 

package/harness-commands/mstar-bootstrap.md

@@ -14,11 +14,10 @@ Distill a coherent knowledge baseline from the current project — useful when t
 
 1. `mstar-harness-core`
 2. `mstar-roles` → `references/project-manager.md`
-3. `skills/pm/SKILL.md` → **§ Host entry** + **§ Boot**
-4. `mstar-plan-conventions`（路径符号）
-5. `mstar-strategy` → **§ STRATEGY.md structure** + **§ Creating STRATEGY.md**
-6. `mstar-compound` → **references/concepts-vocabulary.md**（CONCEPTS.md 规则）
-7. `mstar-compound-refresh` → **§ Core rules**（知识维护基线）
+3. `mstar-plan-conventions`（路径符号）
+4. `mstar-strategy` → **§ STRATEGY.md structure** + **§ Creating STRATEGY.md**
+5. `mstar-compound` → **references/concepts-vocabulary.md**（CONCEPTS.md 规则）
+6. `mstar-compound-refresh` → **§ Core rules**（知识维护基线）
 
 ## Phase 1: Survey — understand what exists
 

package/harness-skills/mstar-coding-behavior/SKILL.md

@@ -202,6 +202,7 @@ Core idea: how you communicate about code matters as much as the code itself.
 
 ## Integration Notes
 
+- **SDD implementer reports** (`mstar-sdd`): completion evidence must include TDD triple — test file(s), command, output — in `task-N-report.md`; fix rounds add the same for new/changed tests.
 - This skill must not be used to bypass:
   - branch constraints,
   - QC/QA gate definitions,

package/harness-skills/mstar-dispatch-gates/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: mstar-dispatch-gates
-description: Morning Star 派发与委派门禁 —— 仅 PM 可增派 subagent、`Execute as` 与 `Delegation`、承接方反递归 NEVER 红线、同条消息 N 次 invoke、QC 三审禁止串行 rollout、Assignment 文案≠派发（invoke 条数须对齐）、未齐不发（emit zero until batch-ready）。**必须**在 `project-manager` 每轮派发、QC 三审并发、双轨 implement、或 leaf 角色疑惑能否 Task 时 Read；所有非 PM 承接方动手前必读反递归与自检。同仓 worktree 与 QC 检出对齐见 `mstar-branch-worktree`。宿主细则见 `mstar-host`（`references/parallel-dispatch.md` 与各宿主 reference）。
+description: Morning Star 派发与委派门禁 —— 仅 PM 可增派 subagent、`Execute as` 与 `Delegation`、承接方反递归 NEVER 红线、SDD implement 串行派发、**SDD 路径 plan QC 强制 tri-review（N=3）**、inline 单席 QC 例外、Assignment 文案≠派发、未齐不发。`project-manager` 派发时必读；leaf 动手前必读反递归。worktree 见 `mstar-branch-worktree`；SDD 见 `mstar-sdd`；宿主见 `mstar-host`。
 ---
 
 ## Load order（必读顺序）
@@ -54,8 +54,9 @@ description: Morning Star 派发与委派门禁 —— 仅 PM 可增派 subagent
 当 PM 声明「并发分派」时，须同时满足**文案并发**与**工具并发**：
 
 - **工具并发**：同一调度轮次内，多个 subagent 调用须在**同一条 assistant 消息**里一次性发出（宿主允许时）。
-- **QC 三审硬门禁（initial wave）**：`qc-specialist` / `qc-specialist-2` / `qc-specialist-3` 须**同一条消息**全部发起（**N=3**）；只发其中一个 = **`dispatch incomplete`**，不得写「三审已并行启动」。
-- **QC targeted re-review**：Assignment 含 **`QC re-review: targeted — reviewers: …`** 时，**N** = 所列席位数（1–3），仍须**同一条消息**发满 **N**；不得默认补满三人。
+- **QC tri-review（SDD 强制）**：`Execution mode: sdd` 且全部 task 完成后 → `qc-specialist` / `qc-specialist-2` / `qc-specialist-3` 同条消息 **N=3**（`qc1`…`qc3` + `qc-consolidated.md`）。Assignment 须含 branch **review-package** 路径。适用于**单 plan 与 iteration**。
+- **QC 单席（例外）**：`Execution mode: inline`（hotfix 等），或 Assignment 显式 `QC mode: single` / `QC mode: single — override: <reason>` → `qc-specialist` ×1，`N=1`，`qc.md`。
+- **QC targeted re-review**：Assignment 含 **`QC re-review: targeted — reviewers: …`** 时，**N** = 所列席位数（1–3），同条消息发满 **N**。
 - **先自检再发送**：发送前核对「Assignment 条数 = 本条消息中的实际 **派发** 调用条数」。
 - **前置步骤与派发回合分离（防串行 rollout）**：为派发准备的 **`bash` / `read` / `glob` / `grep`**（如 `merge-base`、`Review range`、`git rev-parse`）**不计入** `N` 次派发；可在上一条仅含准备的消息完成。准备完成后，**下一条派发消息**须**一次性**含 **`N` 次** Task / subagent invoke。**禁止**先发 `1` 次、等返回再补发其余 `N-1` 次。
 - **未齐不发（emit zero until batch-ready）**：需并发 `N≥2` 而当前只能发 `1` 条时，本条应发 **`0` 条派发 invoke`**（可继续 read/bash 补齐），**禁止**「先发一个顶一下」；`N` 份 payload 就绪后**单次消息发满 `N`**。见 **`mstar-host`** → `references/parallel-dispatch.md`（具备 invoke / Task / subagent 工具的宿主共用）。
@@ -64,11 +65,21 @@ description: Morning Star 派发与委派门禁 —— 仅 PM 可增派 subagent
 
 在支持具名角色 / Task 的宿主上，`## Assignment` **正文不会**拉起子会话。PM 须在**同一条 assistant 消息**（或宿主等价机制）发出与 Assignment **条数一致**的 invoke / Task；仅打印 Markdown = **分派未完成**。**几条 Assignment ⇒ 几次 tool 调用**（默认同消息并行）。
 
+## SDD implement 波次（PM only）
+
+When **`Execution mode: sdd`** (`mstar-sdd`):
+
+- **串行**：one implementer at a time; one **fresh** task reviewer after each — **never** parallel implementers (write conflicts).
+- **`SDD implementer session: sticky`**：same implementer subagent may **resume** across tasks when host supports it; **reviewers never resume** — see **`mstar-sdd/references/sticky-implementer-session.md`**.
+- File handoffs only — no pasted plan/diff/history in dispatch prompts.
+- Record per-task BASE SHA; use `review-package` for diffs — **never `HEAD~1`**.
+- After all tasks: branch `review-package` → **mandatory tri-review N=3** when `Execution mode: sdd`; **N=1** only for `inline` / explicit single override.
+
 ## 并行规则（摘要）
 
-- 独立模块可并行；避免写操作归属重叠；跨领域变更先锁接口契约再并行编码。
-- **QC 三审**在 feature 开发完成后执行；三名 reviewer 共用同一组 `Review cwd` / `Working branch` / `plan_id` / `Review range` / `Diff basis`（**`mstar-branch-worktree`**）。
-- **同一 plan 多 batch**：默认整 plan 交付完成跑一轮完整三审；**fix 后默认 targeted re-review**（N = 被指派的 QC 席位数，同条消息发满 N）；**仅** Assignment 写明 **`QC re-review: full tri-review`** 时复跑三审且用新文件名（**`mstar-plan-artifacts/references/plan-files-and-reports.md`**、**`mstar-review-qc`**）。
+- 独立模块可并行 **implement 轨道**（不同 dev Assignment）；**SDD 单 plan 内 task** 仍串行。
+- **Plan QC tri** after SDD task loop（`Execution mode: sdd`）；**单席**仅 `inline` / hotfix。共用 `Review cwd` / `Working branch` / `plan_id` / `Review range`（**`mstar-branch-worktree`**）。
+- **Tri 同消息规则**：plan QC tri（SDD 或 Assignment 显式 `QC mode: full tri-review`）时三席 **同一条消息**、**同一套** scope 字段。
 
 ## Specialist review-and-edit dispatch
 
@@ -83,8 +94,9 @@ description: Morning Star 派发与委派门禁 —— 仅 PM 可增派 subagent
 
 ## 反模式（派发）
 
-- QC 三审拆在多条消息或等 #1 返回再发 #2/#3。
-- 仅 1 次 invoke 却声称「并行三审已启动」。
+- QC 三审拆在多条消息（tri 模式）或单席却未附 review-package 路径。
+- 仅 1 次 invoke 却声称「tri-review 已并行启动」（tri 模式 N=3）。
+- SDD 并行 implementer dispatch。
 - 递归同角色 subagent；把 Handoff / 多轨编排措辞当 invoke。
 - Review-and-edit 链未完成即 commit integration 分支；PM 代做专业角色编辑而不 invoke。
 - Phase 1 review-and-edit 链三角色并行派发，或未等上一角色返回即派发下一角色。

package/harness-skills/mstar-harness-core/SKILL.md

@@ -30,7 +30,7 @@ description: Morning Star (启明星) harness **强制全局入口** —— 信
 
 ## 最小交付循环
 
-**per-plan**：`specify → clarify → plan` → `plan(locked) → tasks → implement` → QC 三审 + QA → Done。阶段细则 → **`mstar-phase-gates`**。
+**per-plan**：`specify → clarify → plan` → `plan(locked) → tasks → implement`（多 task 默认 SDD）→ plan QC tri + QA → Done（`inline` 单席例外）。阶段细则 → **`mstar-phase-gates`**。
 
 **迭代级**：`iteration-start → [per-plan cycle × N] → iteration-close → PR delivery → PR merge-ready loop`。细则 → **`mstar-iteration`**。
 
@@ -39,11 +39,11 @@ description: Morning Star (启明星) harness **强制全局入口** —— 信
 | 角色 | 始终 | 按任务追加（典型） |
 |------|------|-------------------|
 | **全部** | 本 skill | — |
-| **`@project-manager`** | 本 skill | `mstar-dispatch-gates`、`mstar-phase-gates`、`mstar-plan-conventions`、`mstar-roles`；派 QC 前 `mstar-review-qc`；并行/审查 `mstar-branch-worktree`；plan/status/reports `mstar-plan-artifacts`；UI 类 plan Prepare 阶段 `mstar-design-md`（DESIGN.md 门禁）；新建/大改 skill 时 `mstar-skill-authoring`；迭代管理 `mstar-iteration`（Phase 1–5）；战略性工作 `mstar-strategy`。**不**读 `mstar-coding-behavior` |
+| **`@project-manager`** | 本 skill | `mstar-dispatch-gates`、`mstar-phase-gates`、`mstar-plan-conventions`、`mstar-roles`；implement 波次 `mstar-sdd`；派 QC 前 `mstar-review-qc`；并行/审查 `mstar-branch-worktree`；plan/status/reports `mstar-plan-artifacts`；UI 类 plan Prepare 阶段 `mstar-design-md`（DESIGN.md 门禁）；新建/大改 skill 时 `mstar-skill-authoring`；迭代管理 `mstar-iteration`（Phase 1–5）；战略性工作 `mstar-strategy`。**不**读 `mstar-coding-behavior` |
 | **实现/审查/QA/运维** | 本 skill + `mstar-coding-behavior` + 角色 ref | 有 git 写：`mstar-branch-worktree`；有 plan 路径：`mstar-plan-conventions`（路径符号节）；QC/QA：`mstar-review-qc`；改 status/residual：`mstar-plan-artifacts`；UI 任务：`mstar-design-md`（读取 DESIGN.md tokens）；写入知识库 `{KNOWLEDGE_DIR}`：`mstar-compound`（PM 触发） |
 | **leaf 承接方** | 上栏 + **`mstar-dispatch-gates`**（反递归节） | — |
 
-Routing eval（Cursor 维护）→ `.cursor/skills/mstar-routing-eval/`，**非**运行时必读。
+Routing eval（Cursor 插件内回归用，**非**运行时必读）→ `.cursor/skills/mstar-routing-eval/`。
 
 ## 状态机
 
@@ -88,7 +88,8 @@ PM 在 Assignment 写 **`Task category`**（主类 + 可选 `secondary`）：
 | `mstar-harness-core` | 本文件：入口、状态机、Task category、explore、索引、护栏 |
 | `mstar-phase-gates` | per-plan 双阶段门禁：Prepare/Execute、意图门禁、hotfix、可验证编辑 |
 | `mstar-iteration` | 迭代管理：Phase 1–5（start / Autonomous Execute / iteration-close / PR delivery / PR merge-ready loop） |
-| `mstar-dispatch-gates` | 派发、Delegation、反递归、并行 invoke、QC 同消息三审 |
+| `mstar-dispatch-gates` | 派发、Delegation、反递归、SDD 串行、SDD 路径 plan QC 强制 tri |
+| `mstar-sdd` | Subagent-driven development：file handoff、per-task review、ledger |
 | `mstar-branch-worktree` | 功能分支、worktree、QC/QA 检出对齐 |
 | `mstar-plan-conventions` | `{HARNESS_DIR}` 发现、初始化、Spec 分支模型摘要、产物路径 SSOT |
 | `mstar-plan-artifacts` | 主 plan、reports、`status.json`、residual、knowledge、Done 归档 |
@@ -135,7 +136,8 @@ Read **`mstar-host`** after this skill; detect host per its table, then Read the
 
 | 反模式 | 详见 |
 |--------|------|
-| QC 串行 rollout / dispatch incomplete | `mstar-dispatch-gates` |
+| SDD paste-only / 跳过 task review / 并行 implementer | `mstar-sdd` · `mstar-dispatch-gates` |
+| SDD plan 以单席 `qc.md` 收尾 / 跳过 plan tri | `mstar-dispatch-gates` · `mstar-review-qc` · `mstar-sdd` |
 | fix 后无脑重派三审 / 用 `-rev2` 代替原位复验 | `mstar-plan-artifacts` · `mstar-review-qc` |
 | 递归误派 / 误读 Handoff | `mstar-dispatch-gates` |
 | `quick` 跳过 Prepare | 上表 + `mstar-phase-gates` |

package/harness-skills/mstar-host/SKILL.md

@@ -19,7 +19,7 @@ Read **`mstar-harness-core`** before this skill (even when the host injects proj
 4. Load role via `mstar-roles`
 5. Execute with evidence-first completion checks
 
-Load topic skills **on demand** per `mstar-roles` (do not read every `mstar-*` skill by default). Cursor maint routing-eval: `.cursor/skills/mstar-routing-eval/` only.
+Load topic skills **on demand** per `mstar-roles` (do not read every `mstar-*` skill by default). Cursor routing-eval (`.cursor/skills/mstar-routing-eval/`) is regression tooling only — not part of runtime load order.
 
 ## Detect active host
 

package/harness-skills/mstar-host/references/codex.md

@@ -11,7 +11,7 @@ Parallel PM dispatch: read **`parallel-dispatch.md`** only when Codex exposes an
 - Plugin source: `.codex-plugin/plugin.json`.
 - Runtime skills: repo `skills/` mounted by the Codex plugin (`"skills": "./skills/"`).
 - Custom agent source: repo `codex/agents/*.toml`; CLI/manual install links these into `~/.codex/agents/` or project `.codex/agents/`.
-- `/pm`: shared force entry via the `pm` skill; after that, role behavior comes from `mstar-roles`.
+- **`/pm`** or **`pm` skill**: force PM entry → `mstar-roles` → `project-manager.md` (Codex primary; Cursor/OpenCode for general per-plan work). **`commands/`** when running iteration Phase 1–5.
 - Role files under root `agents/` are for hosts that load OpenCode/Cursor-style agent shells; Codex uses `codex/agents/*.toml` and still loads `mstar-roles` references directly.
 - Tool and plugin availability can be lazy-loaded or session-dependent. Use the tools actually present in the current session; do not infer capability from documentation alone.
 
@@ -38,7 +38,7 @@ Use skill names in prompts and references. Avoid absolute local paths unless the
 - **No invoke tool / no linked custom agent = no dispatch**: printing `## Assignment` does not start another Codex worker.
 - If Codex exposes custom-agent / multi-agent tools and matching Morning Star agents are linked, PM may dispatch through those tools and must follow `parallel-dispatch.md`.
 - If no invoke tool is present when dispatch is required, return **`Blocked`** — report missing invoke capability to the user. Do not substitute single-session role execution in the PM thread unless the user explicitly overrides harness dispatch for this turn.
-- QC tri-review requires three callable reviewer sessions in one dispatch turn. Cannot emit **N=3** → **`Blocked`** for PM rerouting; do not claim tri-review complete or substitute serial/manual review in the PM thread.
+- QC: **`Execution mode: sdd`** → **N=3** tri-review; **`inline`** → **N=1**. Cannot emit required **N** → **`Blocked`**.
 - Leaf executors still follow `mstar-dispatch-gates`: no recursive Task/subagent calls unless Assignment says `Delegation: allowed (...)`.
 
 ## Files, shell, and approvals

package/harness-skills/mstar-host/references/cursor.md

@@ -6,8 +6,8 @@ Parallel PM dispatch: **`parallel-dispatch.md`** (Task tool uses same turn model
 
 ## Cursor-only context
 
-- Role prompts: `mstar-roles`; **`/pm`** → `project-manager` via `pm` skill + `mstar-roles`.
-- Routing-eval: `.cursor/skills/mstar-routing-eval/` — maint only.
+- Role prompts: `mstar-roles`; **`/pm`** or **`pm` skill** → general PM orchestration (per-plan dispatch, gates, QC) without an iteration command. **`commands/`** (`iteration-start`, `iteration-drive`, `mstar-bootstrap`) only when running formal iteration Phase 1–5.
+- Routing-eval: `.cursor/skills/mstar-routing-eval/` — regression tooling only; not runtime load order.
 
 ## Plan mode × harness dual-write
 
@@ -38,13 +38,27 @@ Enforcement: `rules/mstar-cursor-plan-mode.mdc` when plugin active.
 3. `mstar-harness-core` and `mstar-*` skills
 4. `mstar-host` + this reference
 
-## Task tool (QC tri-review)
+## Task tool (QC: SDD → N=3)
 
-- Apply **`parallel-dispatch.md`**: prerequisite message may prep only; dispatch message emits **all `N`** Tasks (**3** for tri-review) in **one** message when parallel is required.
-- `subagent_type`: bare role ids — `qc-specialist`, `qc-specialist-2`, `qc-specialist-3` (same for implement roles: `fullstack-dev`, `frontend-dev`, etc.).
-- Identical across three Tasks: **`plan_id`**, **`Review cwd`**, **`Review range` / Diff basis** (`mstar-branch-worktree`, `mstar-dispatch-gates`, `mstar-review-qc`).
-- Parallel QC ≠ different review cwd per reviewer; one integrated HEAD for scope.
-- Status Update may note QC ran via parallel Task subagents.
+- **`Execution mode: sdd`**: **N=3** Tasks (`qc-specialist`, `qc-specialist-2`, `qc-specialist-3`) + branch review-package path.
+- **`inline`**: **N=1** per `parallel-dispatch.md`.
+- SDD implement/reviewer: **serial** — see **`mstar-sdd`**.
+
+## SDD sticky implementer (Cursor Task resume)
+
+When Assignment has **`SDD implementer session: sticky`** (`mstar-sdd/references/sticky-implementer-session.md`):
+
+| Turn | Task tool |
+|------|-----------|
+| **First task** (or after `fresh` reset) | `subagent_type: <Execute as>` + `implementer-prompt.md` body; capture returned **agent id** → `{SDD_DIR}/implementer-session.json` → `host_agent_id` |
+| **Task 2+** | `resume: <host_agent_id>` + `implementer-continuation-prompt.md` body (new brief path + report path only) |
+
+**Rules:**
+
+- **1 implement dispatch ⇒ 1 Task** per task id (resume counts as the implement dispatch for that task).
+- **Task reviewer**: always **new** Task — **no** `resume` for reviewers.
+- After each task reviewer passes, PM updates `progress.md` and `implementer-session.json` `last_task`.
+- If `resume` fails or agent id missing → reset to **`fresh`** for that task; log reason in Status Update.
 
 ## Dispatch execution（canonical）
 
@@ -68,9 +82,21 @@ Implementation roles use `mstar-coding-behavior` for RCA, test-first checks, rev
 
 ## Gotchas
 
-- Tri-review needs identical `plan_id` and review scope fields.
+- Single-seat and tri-review need identical `plan_id` and review scope fields.
 - Task parallelism does not relax branch/worktree isolation.
 
+## Model tier (SDD + QC)
+
+Map Assignment **`Model tier`** to Task `model` (host-specific slugs):
+
+| Tier | Typical use |
+|------|-------------|
+| `fast` | Transcription tasks; 1–2 file mechanical edits |
+| `standard` | SDD prose implementer; task reviewer floor; plan QC tri seats |
+| `capable` | Large branch QC diff; integration judgment |
+
+**Turn count beats token price** — reviewers and prose implementers use `standard` floor minimum. See `mstar-sdd` SKILL.
+
 ## Project rules
 
 Project `AGENTS.md` / `CLAUDE.md` upward from cwd override global harness defaults when they conflict with Cursor-side rules; user instructions win.

package/harness-skills/mstar-host/references/opencode.md

@@ -6,6 +6,7 @@ Parallel PM dispatch: **`parallel-dispatch.md`** (read in dispatch rounds).
 
 ## Role loading
 
+- **PM entry**: **`/pm`** or **`pm` skill** → `project-manager` for general orchestration; **`commands/`** (`iteration-start`, `iteration-drive`, `mstar-bootstrap`) for formal iteration only.
 - **Role shell**: `agents/<id>.md` referenced by `opencode.json` `agent.<id>` (frontmatter + role binding only).
 - **Role body**: `mstar-roles` `references/<id>.md` (or shared references + parameters).
 - Implementation evidence and RCA behavior: `mstar-coding-behavior`.
@@ -30,6 +31,8 @@ Harness **dispatch** on OpenCode = **one or more `task` tool calls**, each with
 
 PM workflow: finalize Assignment → **call task tool** with **subagent** + generated prompt → wait for subagent Completion Report → update plan / status.
 
+**SDD sticky implementer:** if the task tool exposes **resume** / agent id, follow **`mstar-sdd/references/sticky-implementer-session.md`** and the active host reference. If resume is **not** available, use **micro-batch** (2–3 tasks, one invoke) or **`SDD implementer session: fresh`** per task — do not assume sticky without host support.
+
 ## Role-mention hygiene (OpenCode)
 
 OpenCode may **auto-append** system lines when prompt text stacks multiple agent-id **prefix mentions**. Typical boilerplate (host-generated — **not** harness Assignment):

package/harness-skills/mstar-host/references/parallel-dispatch.md

@@ -27,11 +27,29 @@ Printing `## Assignment` in the main thread **without** matching host invocation
 - Dual-track implement: **`N = 2` ⇒ two invocations in one message** when parallel is required.
 - Status Update on dispatch turns: **`Subagent invokes issued: N`** (must match Assignment count). If Assignments were written but `N = 0` → **`dispatch failed — paste-only`**; fix next message.
 
-## QC tri-review (initial wave)
+## QC default (initial wave)
 
-- Launch `qc-specialist`, `qc-specialist-2`, `qc-specialist-3` in **one** dispatch turn (**N=3** invocations, one message).
-- Do not claim parallel QC unless all three were issued in that turn.
-- Post-dispatch: verify three distinct agent IDs and intended model mapping; on mismatch → invalid dispatch, re-dispatch before consolidation.
+- **`Execution mode: sdd`**: **N=3** tri-review + branch review-package path → `qc1`…`qc3` + consolidated.
+- **`inline` / override**: **N=1** → `qc.md`.
+
+## QC iteration / SDD (same rule)
+
+Formal iteration Phase 2 uses the same SDD + tri rule — not a separate carve-out.
+
+## QC full tri-review（非 SDD 但显式 tri）
+
+当 plan 为 **`Execution mode: inline`** 或单轨，但 Assignment 仍写 **`QC mode: full tri-review`** 时：
+
+- Launch `qc-specialist`, `qc-specialist-2`, `qc-specialist-3` in **one** dispatch turn (**N=3**).
+- Post-dispatch: verify three distinct agent IDs; on mismatch → invalid dispatch.
+
+（**SDD 默认**已在上一节；本节仅覆盖显式 tri 的非 SDD 场景。）
+
+## SDD implement (serial — not parallel)
+
+- **`Execution mode: sdd`**: implementer and task reviewer dispatches are **one at a time** per task.
+- **Never** multiple implementer Tasks in one message for the same plan.
+- See **`mstar-sdd`**.
 
 ## QC targeted re-review (after fixes)
 
@@ -44,5 +62,5 @@ Printing `## Assignment` in the main thread **without** matching host invocation
 1. Required assignments this turn? (`N`)
 2. Prerequisite-only message? → **zero** batch dispatches unless `N = 1`.
 3. Dispatch message contains **exactly `N`** invocation calls?
-4. QC initial tri-review → **exactly 3** in one dispatch message? Targeted re-review → **N** = Assignment reviewer count?
-5. Previous message was prerequisite-only → this message must include **all `N`** (not “QC1 first”).
+4. QC initial: **`Execution mode: sdd`** → **N=3**? **`inline`** → **N=1**? Targeted re-review → **N** = Assignment reviewer count?
+5. SDD implement → **serial** (never batch implementers); sticky = **resume** same implementer, not parallel

package/harness-skills/mstar-iteration/SKILL.md

@@ -1,13 +1,13 @@
 ---
 name: mstar-iteration
-description: Morning Star 迭代管理 —— iteration-start、Autonomous Execute、iteration-close、PR 交付（Phase 4）、PR merge-ready loop（Phase 5）。显式分支策略；compass `{ITERATION_DIR}/`；分支 SSOT：`status.json` metadata + compass frontmatter。
+description: Morning Star 迭代管理 —— iteration-start、Autonomous Execute（per-plan SDD + plan QC tri）、iteration-close、PR 交付（Phase 4）、PR merge-ready loop（Phase 5）。显式分支策略；compass `{ITERATION_DIR}/`；分支 SSOT：`status.json` metadata + compass frontmatter。**必须**在 iteration-start / iteration-drive / 多 plan 编排、integration 分支、iteration-close 或 PR merge-ready 时 Read；`@project-manager` 迭代编排必读。
 ---
 
 # mstar-iteration（迭代管理）
 
 ## Load order
 
-**Read `mstar-harness-core` first.** Path symbols → **`mstar-plan-conventions`**. Per-plan gates → **`mstar-phase-gates`**. Knowledge crystallization → **`mstar-compound`**. On conflict, **`mstar-harness-core` wins**.
+**Read `mstar-harness-core` first.** Path symbols → **`mstar-plan-conventions`**. Per-plan gates → **`mstar-phase-gates`**. Knowledge crystallization → **`mstar-compound`**. **Phase 2 implement 波次**（进入 per-plan implement 前）→ **`mstar-sdd`** + **`mstar-dispatch-gates`**。Phase 2 QC 前 → **`mstar-review-qc`**。On conflict, **`mstar-harness-core` wins**.
 
 ## 设计思路
 
@@ -227,8 +227,18 @@ SSOT = `{HARNESS_DIR}/status.json` + `{PLAN_DIR}/`。todos 只追踪本轮下一
 对每个 active `plan_id`：
 
 1. **Plan start — feature branch**：Assignment 用 `Working branch: create <plan-feature-branch> from <spec_integration_branch>`。一个 plan 一条专用实现分支；内部并行 → topic branches + worktrees（`mstar-branch-worktree`）
-2. **Implement → InReview**：dispatch-only 循环（`§ 2.5`）；每次 Completion Report v2 后更新 `status.json` + 主 plan
-3. **QC → QA → Done**：三审 + QA（`mstar-review-qc`）；gate 全部通过后 PM 标记 `Done`
+2. **Implement → InReview**（`§ 2.5`）：
+   - **默认 `Execution mode: sdd`**（多 task plan；hotfix 可 `inline`）。
+   - PM 载入 **`mstar-sdd`** 后，按 plan task 顺序 **串行** per-task 循环（**不是**一次派发 dev 做全部 tasks）：
+     1. `sdd-workspace <plan-id>` → `{SDD_DIR}`
+     2. `task-brief <plan-file> N` → `{SDD_DIR}/task-N-brief.md`；记录 `BASE_SHA`
+     3. Dispatch **one** implementer subagent（`references/implementer-prompt.md`：brief 路径 + report 路径 + `Model tier`；**禁止**贴整份 plan）
+     4. Implementer `DONE` → `review-package BASE HEAD` → task diff 文件
+     5. Dispatch **one** task reviewer subagent（brief + report + diff + Global Constraints）
+     6. Fix loop 直至 review clean；append `{SDD_DIR}/progress.md`；更新 `status.json` / plan checkbox
+     7. Next task
+   - 每次 Completion Report v2 后更新 `status.json` + 主 plan
+3. **QC → QA → Done**：per-plan 审查链 → **`mstar-sdd`**（L1–L2）+ **`mstar-review-qc/references/review-responsibility-boundaries.md`**（L3 tri / inline 单席）+ QA。
 4. **Plan complete — merge back**：合并 plan feature branch → `spec_integration_branch`；在下一 plan 或 QC 前解决冲突
 5. **Cross-plan 进度同步**：更新 `{ITERATION_DIR}/<iteration-id>-delivery-compass.md` 的 `## Plans` 表状态列
 6. **Next plan** 从步骤 1 继续
@@ -241,21 +251,24 @@ SSOT = `{HARNESS_DIR}/status.json` + `{PLAN_DIR}/`。todos 只追踪本轮下一
 
 ### 2.5 Dispatch-first（implement 派发约束）
 
-**派发回合纪律**（Cursor Task / 具名 invoke 宿主）：
+派发纪律 SSOT → **`mstar-dispatch-gates`** · **`mstar-sdd`** · **`mstar-host/references/parallel-dispatch.md`**。
 
-1. 准备 Assignment 的消息可只含 read/bash — **不计入**派发。
-2. **下一条派发消息**的第一动作 = 发出全部 `Task`（QC 初轮 **N=3** 同条消息；双轨 implement **N=2**）。
-3. `Subagent invokes issued: 0` 而 Assignment 已写出 → **`dispatch incomplete`**；下一条必须补发 invoke，**禁止** PM 线程亲自实现顶替。
+**SDD implement（Phase 2 默认）** — PM **已载入 `mstar-sdd`** 后执行：
 
-| Do | Don't |
-|----|------|
-| **Loop:** `## Assignment` → invoke → Completion Report v2 → 更新 status → next | PM 亲自 Write/Edit/Shell 产品代码 |
-| 1 Assignment ⇒ 1 invoke | Assignment 只写 markdown 不 invoke |
-| merge/branch/handoff 写入 Assignment | 因"上下文已有"而跳过 subagent |
+| 规则 | 说明 |
+|------|------|
+| 串行 | 同一 plan 内 **one implementer at a time**；每 task 后 **one fresh task reviewer** |
+| Sticky（可选） | Assignment **`SDD implementer session: sticky`** + `implementer-session.json`；implementer **resume**，reviewer **fresh** — `mstar-sdd/references/sticky-implementer-session.md` |
+| 文件交接 | brief / report / diff / `progress.md` 在 `{SDD_DIR}`；dispatch prompt **只给路径**，不贴 plan 全文或 task 历史 |
+| Assignment 字段 | 每个 implement dispatch 须含 `Execution mode: sdd`、`SDD dir`、`Model tier`；**禁止**省略 `Model tier` |
+| 大包 inline | **禁止**把 T1–Tn 或整份 plan 写进 **一个** `fullstack-dev` leaf Assignment 冒充 SDD |
+| 分支 diff | 全部 task 完成后 `review-package MERGE_BASE HEAD` → branch diff → plan QC tri（N=3） |
+
+Iteration Phase 2 附加：
 
-- **NEVER** implement while staying PM — 实现一律 delegate dev 角色
-- Delegate scope：`mstar-roles` → PM Execution Boundary
-- 例外：用户显式要求 PM thread 实现；hotfix（`mstar-phase-gates`）
+- PM **NEVER** 在 PM 线程实现产品代码（delegate dev；hotfix 例外见 **`mstar-phase-gates`**）
+- `Subagent invokes issued: 0` 而 Assignment 已写出 → **`dispatch incomplete`**；下一条补发 invoke，禁止 PM 顶替
+- QC 初轮：**SDD → N=3**；**inline → N=1**；双轨 implement → **N=2**（同条消息发满 N）
 
 ### 2.6 Push 纪律
 
@@ -304,26 +317,16 @@ PM **必须**在对话中打印本 checklist；不得默认同过。
 
 ### 3.2 知识结晶（Compound）—— 迭代级核心收口
 
-**Compound 在此执行，不在 per-plan Done 后独立执行。**
-
-PM 触发 `mstar-compound`（可批量）：
-
-1. **收集素材**：回顾本迭代所有 plan 的实现、debug、review 过程，识别以下类型的可结晶知识：
-   - 非平凡 bug 修复及其诊断过程
-   - 新引入的架构模式或约定
-   - 工具链决策及其理由
-   - 跨 plan 重复出现的模式
-   - 有价值的排错经验
-
-2. **逐条判定**：对每条候选知识，回答 `mstar-compound` 的 Q1-Q8 自检；跳过项要在 compass `## Compound Round Summary` 记录原因。
-
-3. **写入或更新**：值得结晶的条目按 `mstar-compound` Phase 1-7 写入 `{KNOWLEDGE_DIR}/<category>/<slug>.md`，或在高重叠时更新已有文档。
+**Compound 在此执行，不在 per-plan Done 后独立执行。** 工作流 SSOT → **`mstar-compound`**（Q1–Q8 自检、Phase 1–7、Phase 6 索引登记强制）。
 
-4. **CONCEPTS.md 协同**：若迭代中引入了新的领域词汇，更新 `<repo-root>/CONCEPTS.md`（`mstar-compound` Phase 5）
+PM 批量触发后须：
 
-5. **索引登记（强制）**：每新增一篇 knowledge doc，必须完成 `mstar-compound` Phase 6，在 `{KNOWLEDGE_DIR}/README.md` 登记一行。Lightweight compound 不豁免 Phase 6。
+1. 收集本迭代 plan 实现 / debug / review 素材，筛候选知识
+2. 逐条过 `mstar-compound` 自检；跳过项记入 compass `## Compound Round Summary`
+3. 写入或更新 `{KNOWLEDGE_DIR}/<category>/<slug>.md`；新领域词更新 `CONCEPTS.md`
+4. **每篇**新 doc 完成 Phase 6（`{KNOWLEDGE_DIR}/README.md` 登记）
 
-若本轮没有值得结晶的知识，仍须在 `## Compound Round Summary` 写明 `无可结晶知识` 及原因。
+若无结晶，仍在 `## Compound Round Summary` 写明 `无可结晶知识` 及原因。
 
 ### 3.3 更新 roadmap
 
@@ -430,7 +433,8 @@ PR **merge** 本身可仍由用户手动执行，除非 Assignment 明确授权
 | `mstar-phase-gates` | per-plan gate 判定 |
 | `mstar-plan-conventions` | 路径符号（`{ITERATION_DIR}`、`{HARNESS_DIR}`） |
 | `mstar-plan-artifacts` | `status.json` SSOT、`{ITERATION_DIR}` 索引维护 |
-| `mstar-review-qc` | QC 三审 — per-plan loop 引用 |
+| `mstar-sdd` | SDD implement 波次 — per-plan loop |
+| `mstar-review-qc` | SDD 强制 plan QC tri — per-plan loop |
 | `mstar-branch-worktree` | 分支/merge/worktree 隔离 |
 | `mstar-compound` | iteration-close 中触发知识结晶 |
 | `mstar-compound-refresh` | iteration-close 后可触发知识维护 |
@@ -440,6 +444,7 @@ PR **merge** 本身可仍由用户手动执行，除非 Assignment 明确授权
 
 - 不要在 per-plan Done 后立即单独 compound——等 iteration-close 统一做
 - 不要在 Autonomous Execute（Phase 2）中修改 per-plan gate 判定
+- **不要在 Phase 2 用 inline 大包 Assignment 替代 SDD per-task 循环**（除非 plan 显式 `Execution mode: inline` / hotfix）
 - 不要用 compass 替代 `status.json` 作为 plan 状态 SSOT
 - 不要在没有完成 per-plan 前置检查的情况下进入 iteration-close
 - 不要在缺少 `iteration_base_branch` / `target_branch` 时默认使用 `main` / `master`
@@ -449,3 +454,4 @@ PR **merge** 本身可仍由用户手动执行，除非 Assignment 明确授权
 - 不要跳过 compound——如果本迭代确实没有可结晶的知识，在 compass `## Compound Round Summary` 写 `无可结晶知识（原因：<简述>）`
 - 不要将 **Phase 4 开 PR** 等同于 **迭代交付完成** — 必须完成 **Phase 5** §5.5 merge-ready loop
 - 不要在 Phase 5 跳过 review comment + resolve 纪律（§5.1 step 5）
+- **不要在 SDD plan 上以单席 `qc.md` 收尾**（除非用户书面 `QC mode: single — override`）

package/harness-skills/mstar-phase-gates/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: mstar-phase-gates
-description: Morning Star (启明星) Spec-Driven 双阶段门禁 —— Prepare（`specify → clarify → plan`）、Execute（`plan(locked) → tasks → implement`）、意图门禁、长期目标优先、分批 roadmap 强制落盘、clarify 核心纪律（共享理解 / 先探索 / 每问推荐答案）、hotfix 压缩路径、可验证编辑、Phase Gate 最小证据。**必须**在 PM 判定 gate、首次 implement 派单前、产品/架构参与 Prepare、或解释为何不能跳过 plan/clarify 时 Read；`@project-manager` 每轮编排非 hotfix 任务必读；`@product-manager` / `@architect` 写规格与锁 plan 时必读 Prepare 节；实现角色 Read Execute 与 hotfix 例外即可。Task category 与 `quick` 禁豁免规则仍在 `mstar-harness-core`。实现行为见 `mstar-coding-behavior`。迭代级活动（iteration-start / iteration-close / compound）见 `mstar-iteration`。
+description: Morning Star (启明星) Spec-Driven 双阶段门禁 —— Prepare（`specify → clarify → plan`）、Execute（`plan(locked) → tasks → implement`；多 task 默认 **`Execution mode: sdd`**）、意图门禁、长期目标优先、分批 roadmap 强制落盘、clarify 核心纪律、hotfix 压缩路径、可验证编辑、Phase Gate 最小证据。**必须**在 PM 判定 gate、首次 implement 派单前、产品/架构参与 Prepare、或解释为何不能跳过 plan/clarify 时 Read；`@project-manager` 每轮编排非 hotfix 任务必读；`@product-manager` / `@architect` 写规格与锁 plan 时必读 Prepare 节；实现角色 Read Execute 与 hotfix 例外即可。Task category 与 `quick` 禁豁免规则仍在 `mstar-harness-core`。实现行为见 `mstar-coding-behavior`；SDD 见 `mstar-sdd`；迭代级活动见 `mstar-iteration`。
 ---
 
 ## Load order（必读顺序）
@@ -29,7 +29,7 @@ description: Morning Star (启明星) Spec-Driven 双阶段门禁 —— Prepare
 - **`plan(locked)`** — 冻结基线；实现中出现新约束时**先回写 plan 再继续**。
 - **`tasks`** — 含依赖顺序、并行标记、完成判据；每任务可追踪到 plan、roadmap 批次与验收标准。
   - **并行标签**：若 PM 将 ≥2 条实现轨 **同时** 分派，须在 Assignment 写明 `Dispatch mode: parallel independent tracks`；同仓 ≥2 可写并发时写明 `Worktree isolation: required`（见 **`mstar-dispatch-gates`** 与 **`mstar-branch-worktree`**）。
-- **`implement`** — 按 tasks 顺序执行并提交自检证据；完成进入 `InReview`；遵循 **`mstar-coding-behavior`**。
+- **`implement`** — 按 tasks 顺序执行；多 task plan **默认** `Execution mode: sdd`（`mstar-sdd`）；hotfix 可 `inline`。完成进入 `InReview`；遵循 **`mstar-coding-behavior`**。
 
 ### 可验证编辑与上下文纪律
 
@@ -39,55 +39,20 @@ description: Morning Star (启明星) Spec-Driven 双阶段门禁 —— Prepare
 
 ### Hotfix 例外
 
-- 压缩为 `specify(min) → plan(min) → implement`；必须在回报或 plan notes 补记事后 **`clarify/RCA`**。
+压缩路径与事后补记见下文 **§ Hotfix 例外**（Playbook 末尾）。
 
 ## Phase Gate Playbook
 
-本手册将 `specify -> clarify -> plan -> tasks -> implement` 变成可执行动作，供 `@project-manager`、开发与 QA 在日常交付中快速对齐。
+执行动作与最小产物见上文 **§ Spec-Driven 双阶段门禁**。本节仅 **Playbook 补充**（不重复 Prepare/Execute 长文）。
 
-## 适用范围
+### Execute 补充（Playbook 专有）
 
-- 非热修（non-hotfix）任务默认强制执行全链路门禁。
-- 热修可走压缩路径，但必须补事后 `clarify/RCA` 记录。
-
-## 两阶段门禁（per-plan Playbook）
-
-### A. Prepare
-
-顺序：`specify -> clarify -> plan`
-
-- `specify`
-  - 目标：定义问题、范围、验收。
-  - 最小产物：问题陈述、目标用户价值、非目标、DoD 草案。
-- `clarify`
-  - 目标：收敛会影响方案或验收的歧义。
-  - 最小产物：歧义清单 + 结论；若未收敛则 `blocked`。
-  - **意图**：区分字面请求与真实目标；手段/目标混淆须在此收敛（见本 skill `SKILL.md` Intent gate）。
-  - **结构化澄清**：与用户核对歧义或决策时，`@project-manager`（及直接与用户对话的角色）在**宿主支持**时优先用 `question` 类能力拉齐输入；否则用等价结构化正文。宿主差异细则见当前宿主的 `mstar-host` skill。
-  - **`clarify` 核心纪律**（见 **`mstar-phase-gates` SKILL.md** Prepare · `clarify`）：逐方面核对至共享理解；沿设计决策树逐枝、一次一决；能探索代码库则先探索；每问附推荐答案；阶段末汇总已决与仍 open 假设。
-- `plan`
-  - 目标：给出可执行技术方案与风险控制。
-  - 最小产物：目标状态、方案、模块边界/接口契约、风险与回滚、验证计划。
-  - **准入**：能书面写出真实目标、成功判据、非目标后再锁 plan（同 `SKILL.md`）。
-  - **分批准入**：若计划不是一次性交付完整范围，须写 `Roadmap / Batch Plan`：本批做什么、后续批次做什么、依赖顺序、暂缓项、owner/触发条件、最终 Done 定义；缺失则 `blocked`。
-
-### B. Execute
-
-顺序：`plan locked -> tasks -> implement`
-
-- `plan locked`
-  - 目标：冻结本轮基线，防止边做边漂移。
+- **`plan locked`**
   - 最小动作：在 plan 或 notes 记录当前锁定版本（日期或 hash）。
-- `tasks`
-  - 目标：把 plan 拆成可执行任务与依赖顺序。
-  - 最小产物：任务列表、并行标记、完成判据、映射到验收标准与 roadmap 批次。
-  - **PM**：若并行标记对应「多轨同时 implement」，在对外 **Status Update** 与实现 Assignment 写入 **`Dispatch mode: parallel independent tracks`**；同仓多可写并发时叠 **`Worktree isolation: required`**（见 `mstar-dispatch-gates` 与 `mstar-branch-worktree`）。
-- `implement`
-  - 目标：按任务执行并提交证据，进入审查。
-  - 最小产物：实现 diff、自检证据、回报与 handoff。
-  - **行为准则**：执行中遵循 `mstar-coding-behavior`（不静默假设、优先简单方案、只做与任务直接相关的手术式改动、按 `Step -> verify` 推进）。
-  - **编辑纪律**：改文件前以磁盘为准重读；Patch 失败则重读、缩小步长，禁止盲试（见 **`mstar-phase-gates` SKILL.md**「可验证编辑与上下文纪律」）。
-  - **知识库 / 迭代 compass**：若项目在 `plans[].metadata` 中登记了 `primary_spec` / `spec_refs` / `iteration_compass` / `iteration_refs`，**开工前**须阅读并在回报中说明已对齐；规则见 **`mstar-plan-conventions`** 与 **`mstar-plan-artifacts/references/knowledge-and-designs.md`**。
+  - **Plan 质量门**（新 plan / 大改）：无 placeholder（`...`、`TBD`、`etc.`）；含 **Global Constraints** 与 per-task **Interfaces**；PM self-review 三问（每 task 可独立验证？依赖顺序清晰？无隐含假设？）— 见 `mstar-plan-artifacts/templates/plan.main.md`。
+- **`implement`**
+  - 最小产物：实现 diff、自检证据、回报与 handoff；行为准则 → **`mstar-coding-behavior`**；编辑纪律 → 上文「可验证编辑与上下文纪律」。
+  - **知识库 / 迭代 compass**：若 `plans[].metadata` 登记了 `primary_spec` / `spec_refs` / `iteration_compass` / `iteration_refs`，**开工前**须阅读并在回报中说明已对齐 → **`mstar-plan-conventions`** · **`mstar-plan-artifacts/references/knowledge-and-designs.md`**。
 
 ## 角色职责
 
@@ -112,7 +77,7 @@ per-plan Done 是 per-plan 的闭环终点；compound 是迭代级收口活动
 
 ## Plan 目录与审查报告（启用 `{PLAN_DIR}` 时）
 
-- 进入 `InReview` 后，QC 书面产出落入 `{PLAN_DIR}/reports/<plan-id>/`（如 `qc1.md` … `qc-consolidated.md`）；**fix 后默认 targeted re-review**（原位更新同文件，不默认 `-rev2`），见 **`mstar-plan-artifacts/references/plan-files-and-reports.md`**。**多 batch**：完整三审**默认在整 plan dev 完成后一次**（非每 batch）。
+- 进入 `InReview` 后，QC 产出落入 `{PLAN_DIR}/reports/<plan-id>/`（**SDD 默认 tri** `qc1`…`qc-consolidated`；**inline** 单席 `qc.md`）。**fix 后默认 targeted re-review**。SDD per-task review 在 implement 波次内完成（`mstar-sdd` task reviewer）。
 - 非阻断项与后续技术债：PM 汇总后写入 `{HARNESS_DIR}/status.json` 根级 `residual_findings[<plan-id>]`（**open**，与 `plans` 平级；canonical 见 **`mstar-plan-artifacts` SKILL.md**）；关闭后迁入 `{HARNESS_DIR}/archived/residuals/<plan-id>.json`，与 `mstar-review-qc` 一致。每条 **`severity`** 遵守 **`mstar-plan-artifacts/references/status-and-residuals.md`**「Residual findings：severity（SSOT，机器字段）」。
 
 ## 快速判定（PM）

package/harness-skills/mstar-plan-artifacts/SKILL.md

@@ -12,6 +12,7 @@ description: Morning Star plan harness artifacts — `{PLAN_DIR}` main plans and
 | Topic | See |
 |-------|-----|
 | Main plan, reports naming, QC waves, residual and plan index order | `references/plan-files-and-reports.md` |
+| Plan template (Global Constraints, Interfaces) | `templates/plan.main.md` |
 | knowledge / iterations / specs boundaries and indexes | `references/knowledge-and-designs.md` |
 | Done row compaction Profile A/B | `references/done-compaction.md` |
 | `status.json`, residual severity, lifecycle, `jq` | `references/status-and-residuals.md` |