@mstar-harness/opencode 0.7.5 → 0.7.7

package/harness-commands/iteration-drive.md

@@ -6,7 +6,34 @@ agent: project-manager
 
 # Drive Iteration
 
-Drive the active Morning Star iteration forward. The canonical flow is in **`mstar-iteration`** — this command loads it and executes its phases.
+Drive the active Morning Star iteration forward. **Boot loads skills; this command sequences Phase 2 → Phase 3 → PR.** Domain SSOT → **`mstar-iteration`**.
+
+## Phase state machine（禁止跳步）
+
+```
+Phase 2: Autonomous Execute  →  Phase 3: iteration-close  →  Completion: PR
+   (per-plan dispatch loop)        (独立 gate，非 final plan)      (close 之后)
+```
+
+| 当前状态 | 允许 | 禁止 |
+|----------|------|------|
+| 存在非 `Done` plan | Phase 2：dispatch implement / QC / QA | 进入 Phase 3 或开 PR |
+| **全部** plan `Done` | **立刻** `## Phase 3: iteration-close`（§3.0→§3.5） | 宣称 iteration 完成、写 PR、结束会话 |
+| Phase 3 exit checklist 全 `[x]` + compass `status: completed` | `git commit` → push → PR | 跳过 compound / roadmap / frontmatter |
+
+**常见误判**：final plan 的 closure notes、compound 草稿、roadmap 口头更新 **≠** iteration-close。必须打印 §3.1 entry + §3.5 exit checklist。
+
+## PM invariants（Phase 2 全程有效）
+
+| 禁止（PM 线程） | 必须 |
+|-----------------|------|
+| Write/Edit/Shell 产品代码、写测试、跑 QC 审查 | 每条 implement/QC/QA Assignment ⇒ **1 次 `Task`** |
+| 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 PR / 汇报结束 | **先** Phase 3（`mstar-iteration` §3），**再** PR（见 Completion） |
+
+派发细则 → **`mstar-dispatch-gates`** + **`mstar-host`**。Phase 3 细则 → **`mstar-iteration` §3** + **`mstar-compound`**。
+
+**Session todos**：进入 Phase 2 时设 plan-wave todos；当仅剩 1 个非 `Done` plan 时 **追加** `phase-3-iteration-close`（`mstar-iteration` §2.1）；§3.5 完成前不得勾掉。
 
 ## Boot
 
@@ -42,13 +69,18 @@ Execute **`mstar-iteration` § Phase 2** exactly. Summary:
    - Merge plan branch → integration branch
    - Cross-plan progress sync → compass
    - Next plan
-6. Repeat until **all** plans `Done` → exit loop, enter Phase 3
+6. Repeat until **all** plans `Done` → **STOP**（见 Phase state machine）→ 打印 `## Phase 3: iteration-close` → 执行 **`mstar-iteration` §3**；**不得**开 PR
 
-**Dispatch-first constraint** (§ 2.5): PM never implements directly; `1 Assignment ⇒ 1 invoke`.
+派发回合纪律 → **`mstar-dispatch-gates`** + **`mstar-iteration` §2.5**。
 
 ## Phase 3: iteration-close
 
-当 **every** plan 为 `Done` 时，**STOP** per-plan loop，打印 `## Phase 3: iteration-close`，按 **`mstar-iteration` § Phase 3** 逐步执行。final plan 的 closure 只能提供输入，不能替代 close gate。
+当 **every** plan 为 `Done` 时：
+
+1. **STOP** per-plan loop — 不得合并 plan 分支、写 PR、或输出「迭代完成」摘要。
+2. 打印标题 **`## Phase 3: iteration-close`**（用户可见的 phase 边界）。
+3. 按 **`mstar-iteration` § Phase 3** 逐步执行 §3.0→§3.5。final plan 的 closure 只能提供输入，不能替代 close gate。
+4. §3.5 exit checklist 全 `[x]` 且 compass frontmatter `status: completed` 后，才进入下方 **Completion: Create PR**。
 
 | Step | Section | 要点 |
 |------|---------|------|
@@ -69,6 +101,8 @@ git push origin <spec_integration_branch>
 
 ## Completion: Create PR
 
+**Precondition（HARD）**: Phase 3 complete — §3.5 exit checklist 全 `[x]`；compass YAML `status: completed` + `end_date`；compound Phase 6 索引已登记（或已记录无可结晶原因）。**不满足 → 禁止开 PR。**
+
 All iteration-close changes committed to integration branch:
 
 - Resolve PR target: `status.json` → `metadata.target_branch`（fallback：compass frontmatter `target_branch`）

package/harness-commands/iteration-start.md

@@ -1,6 +1,6 @@
 ---
 name: iteration-start
-description: Start a new harness iteration — research backlog, lock direction with grill-me, produce compass/plans, run mandatory Review & Edit chain via Task dispatch (product-manager → architect → writing-specialist each review-and-edit, then PM final lock), then create the integration branch from an explicit base. Not Done until review chain completes, compass is locked, and base/target branches are recorded.
+description: Start a new harness iteration — research backlog, lock direction with grill-me, produce compass/plans, run mandatory Review & Edit chain via sequential Task dispatch (product-manager → architect → writing-specialist, each review-and-edit, then PM final lock), then create the integration branch from an explicit base. Not Done until review chain completes, compass is locked, and base/target branches are recorded.
 agent: project-manager
 ---
 
@@ -8,6 +8,20 @@ agent: project-manager
 
 Start a new Morning Star harness iteration. **Not Done until the Review & Edit chain runs via dispatched roles and PM lock — not when compass files are first written.**
 
+## PM invariants（本命令全程有效 — 读完再动手）
+
+你是 **`project-manager` 编排者**，不是三专业角色的合并替身。
+
+| 禁止（PM 线程） | 必须（宿主有 Task 时） |
+|-----------------|------------------------|
+| 自己 Edit compass/plans/specs 冒充 @product-manager / @architect / @writing-specialist 的审查编辑 | §5.1 → §5.2 → §5.3 **顺序**各 **1 次 invoke**；上一角色返回后再派发下一角色 |
+| 只写 `## Assignment` 或 checklist 就声称 review chain 完成 | **几条角色 ⇒ 几条 invoke**；零 invoke = `dispatch incomplete`（`mstar-dispatch-gates`） |
+| §5 完成前 commit / 创建 integration 分支 | 5.4 PM lock 在 subagent 返回且磁盘产物已修订之后（`mstar-iteration` §1.6） |
+
+派发细则 → **`mstar-dispatch-gates`**（specialist review-and-edit dispatch）+ **`mstar-host`**（宿主 invoke 能力）。**不得**在 PM 线程加载其他 role reference 代劳。
+
+**完成定义**：compass `status: locked` + 三角色 invoke 已返回 + pre-commit checklist 全 `[x]` — 不是初稿落盘。
+
 Detailed workflow → **`mstar-iteration` § Phase 1: iteration-start**（含 §1.6 Review & Edit chain）；per-plan Prepare gates → **`mstar-phase-gates`**（specify → clarify → plan）。
 
 ## 0. Boot
@@ -16,9 +30,10 @@ Detailed workflow → **`mstar-iteration` § Phase 1: iteration-start**（含 §
 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` → iteration-start review chain + implement dispatch
+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）
 
 ## 1. Research
 
@@ -49,9 +64,13 @@ Explore candidate directions targeting **product completeness**:
 - Allow substantive refactoring where it accelerates product maturity
 - Scope 2–4 candidates, each with clear product-completeness goals and trade-offs
 
-## 3. Lock Direction — grill-me
+## 3. Lock Direction — bundled `grill-me`
 
-Run the **grill-me skill** to stress-test candidate directions with the user:
+This command bundles a **non-`mstar-*`** skill at `skills/grill-me/SKILL.md`. **Only this command step** references it — **do not** load it from `mstar-harness-core` or other `mstar-*` skills.
+
+**Before this step:** Read `skills/grill-me/SKILL.md`.
+
+Run **grill-me** to stress-test candidate directions with the user:
 
 - Walk through trade-offs for each candidate
 - Converge on a **single iteration direction** with shared understanding
@@ -74,20 +93,27 @@ Produce harness artifacts per **`mstar-iteration` § 1.3**（template: `mstar-it
 
 **STOP**: Do not run §6 Integration Branch until **all** rows below are true.
 
+**顺序（HARD）**：`product-manager` → `architect` → `writing-specialist` → PM lock。后一角色基于前一角色已落盘的修订继续编辑；**禁止**三角色并行 invoke。
+
 Each role below **reviews and directly edits** the documents. Do not just flag issues — apply the fixes yourself. PM only steps in for the final lock.
 
-| # | Role | Required action |
-|---|------|-----------------|
-| 5.1 | **@product-manager** | Task subagent: **edit** compass, plans, `{SPECS_DIR}/`; scope, UX, priorities |
-| 5.2 | **@architect** | Task subagent: **edit** compass, plans, specs; contracts, SQL, module boundaries |
-| 5.3 | **@writing-specialist** | Task subagent: **edit** all iteration docs; terminology, structure, clarity |
-| 5.4 | **@project-manager** | Merge subagent edits; resolve conflicts; **lock** compass (`status: locked`); confirm Prepare gates |
+| # | Role | Required action | Gate |
+|---|------|-----------------|------|
+| 5.1 | **@product-manager** | invoke: **edit** compass, plans, `{SPECS_DIR}/`; scope, UX, priorities | 完成后方可 5.2 |
+| 5.2 | **@architect** | invoke: **edit** compass, plans, specs（含 5.1 修订后版本）; contracts, SQL, module boundaries | 5.1 返回后；完成后方可 5.3 |
+| 5.3 | **@writing-specialist** | invoke: **edit** all iteration docs（含 5.1–5.2 修订后版本）; terminology, structure, clarity | 5.2 返回后 |
+| 5.4 | **@project-manager** | Merge subagent edits; resolve conflicts; **lock** compass (`status: locked`); confirm Prepare gates | 5.3 返回后 |
 
 **Evidence of done** = edited compass / plans / specs on disk + compass `status: locked`. **No** separate iteration review reports under `reports/` — unlike per-plan QC, there is no downstream audit chain to preserve.
 
-**Tool rule (Cursor / hosts with Task)**:
-- Steps 5.1–5.3: **MUST** use `Task` (parallel when independent). See **`mstar-dispatch-gates`** · iteration-start review chain.
-- PM thread **MUST NOT** substitute by performing all three specialist edits itself.
+**Tool rule (hosts with invoke)** — per **`mstar-dispatch-gates`** · specialist review-and-edit dispatch（**顺序链**，非 parallel batch）：
+
+1. 为当前角色写好 Assignment（含须直接编辑的文件路径；5.2/5.3 注明基于上一角色已落盘修订）。
+2. **派发 1 次 invoke** → 等待 Completion Report / 磁盘修订完成。
+3. 按序重复下一角色：**5.1 → 5.2 → 5.3**（每步一条派发消息、一次 invoke）。
+4. 本条链 **禁止** PM 线程代做专业编辑；**禁止** 同条消息并行三角色；**禁止** 未等 5.1 返回就发 5.2/5.3。
+5. 5.3 返回后，PM 线程再做 5.4 merge + lock。
+
 - Exception: user explicitly waives subagent dispatch ("PM-only review").
 
 **Prepare gate (per plan in compass)**:

package/harness-skills/grill-me/SKILL.md

@@ -0,0 +1,10 @@
+---
+name: grill-me
+description: Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".
+---
+
+Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.
+
+Ask the questions one at a time.
+
+If a question can be answered by exploring the codebase, explore the codebase instead.

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

@@ -26,7 +26,7 @@ description: DESIGN.md design system specification for Morning Star projects. Cr
 
 **YAML frontmatter is the SSOT for token values.** Colors, typography, spacing, rounded, and components live in the frontmatter as structured, machine-readable data. The Markdown body is supplementary documentation (rules, intent, usage guidance). When reading DESIGN.md, always parse the YAML frontmatter first; when writing, keep frontmatter and body in sync.
 
-**Out of scope:** rendered UI artifacts (use Open Design MCP `user-open-design`); frontend implementation that consumes DESIGN.md tokens (use `@frontend-dev` / `@fullstack-dev`); QC review verdict rules (→ **`mstar-review-qc`**).
+**Out of scope:** rendered UI preview tooling outside the repo; frontend implementation that consumes DESIGN.md tokens (use `@frontend-dev` / `@fullstack-dev`); QC review verdict rules (→ **`mstar-review-qc`**).
 
 ## Location
 
@@ -129,14 +129,6 @@ gray-1000: #000   gray-1000: #eee
 - Token names are the **SSOT interface** — consumers reference tokens by name, not raw values
 - `references/design-md-spec.md` § Light/Dark rules defines the contract
 
-## Open Design integration
-
-If `user-open-design` MCP is available:
-- DESIGN.md is the **design intent** (tokens + rules)
-- Open Design manages **rendered artifacts** (HTML/CSS/JSX previews)
-- When DESIGN.md is updated, recommend syncing the Open Design project to reflect new tokens
-- DESIGN.md does not replace Open Design; they are complementary
-
 ## References
 
 - `references/design-md-spec.md` — normative spec: section definitions, token naming conventions, light/dark contract

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

@@ -70,20 +70,26 @@ description: Morning Star 派发与委派门禁 —— 仅 PM 可增派 subagent
 - **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`**）。
 
-## iteration-start Review & Edit chain
+## Specialist review-and-edit dispatch
 
-与 QC 三审同理：**dispatch-first** 适用于文档审查角色，不限于 implement/QC。
+当 PM 派发**文档编辑类**专业角色（如 product-manager、architect、writing-specialist）直接修订 harness 产物时：
 
-- PM 写 compass/plans **初稿**；**@product-manager**、**@architect**、**@writing-specialist** 各通过 Task **直接编辑** compass / plans / specs（**不**另写 `reports/` 审查报告）。
-- PM 线程跳过派发、自行完成三角色全部编辑 = **反模式**（见 **`mstar-harness-core`** 反模式索引、`mstar-iteration` §1.6）。
-- Steps 5.1–5.3 须 **Task**（可并行）；PM lock（5.4）在 PM 线程完成。**不得**在 review chain 完成前 commit integration 分支。
+- PM 写初稿；各角色通过宿主 invoke **直接编辑**目标文件（**不**另写仅评论式 `reports/` 替代修订）。
+- **1 Assignment ⇒ 1 invoke**。
+- **Phase 1 Review & Edit chain**（`mstar-iteration` §1.6）：**顺序** `product-manager` → `architect` → `writing-specialist`；上一角色落盘修订完成后再派发下一角色。**禁止**本链并行 batch（与 QC 三审不同）。
+- 其他彼此独立、无先后依赖的文档编辑任务：可并行（同条消息发满 N），见 **`parallel-dispatch.md`**。
+- PM 线程代做全部专业编辑 = **反模式**（`mstar-iteration` §1.6、`mstar-harness-core` 反模式索引）。
+- PM merge / lock（如 compass `status: locked`）在链末 subagent 返回后于 PM 线程完成。**不得**在 review-and-edit 链完成前 commit integration 分支。
 
 ## 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」。
 - 递归同角色 subagent；把 Handoff / 多轨编排措辞当 invoke。
-- iteration-start 在 Review & Edit chain 完成前 commit；PM 代做三角色全部编辑而不派发 Task。
+- 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。
+- 全部 plan `Done` 后跳过 Phase 3 直接 PR；final plan closure 替代 `mstar-iteration` §3.1–§3.5。
+- Assignment 已写、invoke 为零（paste-only）却进入下一 gate。
 
 ## References
 

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

@@ -11,7 +11,13 @@ description: Morning Star (启明星) harness **强制全局入口** —— 信
 
 - 凡 **`mstar-*`**（`name` ≠ `mstar-harness-core`）假定读者**已 Read 本 skill**。
 - **仅读专题、未读核心** → 未完成 harness 加载。
-- 各专题 SKILL.md 含 **Load order**；按 **`mstar-roles`** 角色矩阵按需加载，**禁止**为「保险」通读全部专题。
+- 各专题 SKILL.md 含 **Load order**；按 **`mstar-harness-core`** 专题 skill 索引 + 角色 load contract 按需加载，**禁止**为「保险」通读全部专题。
+
+## Standalone harness（`mstar-*` 自洽）
+
+- **`mstar-*` skill 体系**在本仓库内自洽运行：正文与 load order **不得**依赖仓库外的 skills、CLI、或 MCP 服务。
+- **Commands**（如 `/iteration-start`）可引用本仓库 bundled、**非 `mstar-*`** 的辅助资产（例如 `skills/grill-me/`）；该引用**仅**存在于 command 层，**不**进入 `mstar-*` 专题索引或 load matrix。
+- 框架/SDK/API 问题：先 Read/Grep 项目内文档与源码；不将第三方文档工具写入 `mstar-*` 必读路径。
 
 ## 信息源优先级
 
@@ -33,7 +39,7 @@ 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`（iteration-start/drive/close 含 compound）；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`；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 1–3 含 compound）；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） | — |
 
@@ -117,9 +123,9 @@ Read **`mstar-host`** after this skill; detect host per its table, then Read the
 - plan-writing artifacts land in `{PLAN_DIR}`, not external default plan directories.
 - PM Assignment 键名英文；任务正文可中文；产出/报告默认英文。
 
-## 库文档（Context7）
+## 库文档 / API 问题
 
-框架/SDK/API 问题 → **`references/library-docs-protocol.md`**（MCP 优先，CLI 兜底，禁双跑）。
+先 Read/Grep 项目内文档、规格与源码；仍不确定时再向用户澄清。**不**将第三方 MCP/CLI 文档工具写入 `mstar-*` load order。
 
 ## 升级触发
 
@@ -139,19 +145,22 @@ Read **`mstar-host`** after this skill; detect host per its table, then Read the
 | 无证据宣称完成 | `mstar-coding-behavior` / verification |
 | CreatePlan 不落盘 / 无 `{HARNESS_DIR}` mirror | `mstar-host` · `cursor-plan-mode-bridge` |
 | 临时方案 / 后续计划只写在对话里 | `mstar-phase-gates` · Durable Roadmap Gate |
-| iteration-start commit before review chain | `mstar-iteration` §1.6；PM 模拟三角色编辑而不派发 Task |
-| Phase 3 折叠进 final plan closure / 跳过 §3.1 gate | `mstar-iteration` §3.0–§3.5；`iteration-drive` Phase 3 |
+| Phase 1 review chain 未完成即 commit | `mstar-iteration` §1.6；PM 代做专业角色编辑；或三角色并行派发 |
+| Phase 2 paste-only / PM 自实现 | `mstar-dispatch-gates` |
+| Phase 3 折叠进 final plan closure / 跳过 §3.1 gate | `mstar-iteration` §3.0–§3.5 |
 | iteration-close 无 frontmatter completed / 漏 compound Phase 6 | `mstar-iteration` §3.0.5、§3.4、§3.2 #5 |
 | iteration 默认 `main` 作 base 或 PR 目标 | `mstar-iteration` §1.2、§2.3；`mstar-plan-conventions` Spec 分支模型 |
 
-## 可选：OpenViking Memory
-
-会话有 **`memsearch`** 工具时 Read **`references/openviking-memory-plugin.md`**。
+## 常见 harness 说法对照（帮助理解角色分工）
 
-## References（保留在 core）
+| 常见说法 | 本仓库实体 |
+|----------|------------|
+| 总编排 | `@project-manager` |
+| 规划/访谈 | `@product-manager` / `@architect` + Prepare 阶段 |
+| category 路由 | **`Task category`** + 路由表 + 子代理选择 |
+| 持续推进 / 不半途而废 | Phase Gate + Todo/`tasks` + 验证门禁 |
+| 行级哈希锚定编辑 | 以 **读后再改 + 小步 Patch** 纪律落实（`mstar-phase-gates`） |
 
-- `references/open-harness-principles.md` — 理念与分层 `AGENTS.md`
-- `references/library-docs-protocol.md` — Context7
-- `references/openviking-memory-plugin.md` — 条件加载
+项目根 **`AGENTS.md` 写什么、分层 bootstrap** → **`mstar-plan-conventions`** `references/harness-bootstrap-and-agents-layering.md`（勿在 core 重复长文）。
 
 **专题 skill**（规则在各自 `SKILL.md`）：`mstar-phase-gates`、`mstar-branch-worktree`、`mstar-plan-artifacts` 等 — 见上表索引。

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

@@ -35,11 +35,7 @@ Use **capability signals** (not filesystem paths):
 
 ## Parallel dispatch (invoke-capable hosts)
 
-When PM dispatches **N >= 2** concurrent assignees (QC tri-review, dual-track implement, etc.) and the host exposes actual invoke / Task / subagent tools, read **`references/parallel-dispatch.md`** in the dispatch round (shared with `mstar-dispatch-gates`). Without a callable invoke tool, Assignment Markdown is not dispatch.
-
-## Library docs (Context7)
-
-Follow `mstar-harness-core` → `references/library-docs-protocol.md`.
+When PM dispatches **N >= 2** concurrent assignees (QC tri-review, dual-track implement, etc.) and the host exposes actual invoke / Task / subagent tools, read **`references/parallel-dispatch.md`** in the dispatch round (shared with `mstar-dispatch-gates`). Without a callable invoke tool when dispatch is required → **`Blocked`**; Assignment Markdown alone is not dispatch.
 
 ## Conflict order
 

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

@@ -37,8 +37,8 @@ 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, use single-session role execution: state the active role, load that role reference, complete the assignment, and return Completion Report v2.
-- QC tri-review is only "parallel" when three distinct callable reviewer sessions are actually launched in one dispatch turn. Without that, run a clearly labeled serial/manual review path or return `Blocked` for PM rerouting.
+- 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.
 - 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

@@ -46,15 +46,16 @@ Enforcement: `rules/mstar-cursor-plan-mode.mdc` when plugin active.
 - Parallel QC ≠ different review cwd per reviewer; one integrated HEAD for scope.
 - Status Update may note QC ran via parallel Task subagents.
 
-## Supplemental execution modes
+## Dispatch execution（canonical）
 
-**Mode A — Single session, multiple roles:** PM writes Assignment; executor states `Acting as role: …` and loads `mstar-roles` reference. QC/QA without Task: same field alignment rules.
+Cursor PM dispatch = **`Task`** with `subagent_type` matching the Assignment `Execute as` role.
 
-**Mode B — Multi-window:** One Assignment per chat; first turn loads role via `mstar-roles`; PM consolidates in main thread.
-
-**Mode C — Worktrees:** `mstar-branch-worktree` + `mstar-dispatch-gates` for concurrent writers; do not assign different tri-review cwd by default.
-
-**No recursive Task in implement subagents:** Task recipient is already `Execute as`; must not re-spawn same dev role. Assignment wins over conflicting outer messages (`Delegation: forbidden`).
+- **1 Assignment ⇒ 1 Task**; parallel batches ⇒ **N Tasks in one message** (`parallel-dispatch.md`, `mstar-dispatch-gates`).
+- Assignment Markdown **does not** start work. PM thread **must not** implement, review, or edit specialist deliverables by loading another role reference in the same session (`Acting as role: …` is **not** dispatch).
+- No callable `Task` / subagent for required work → **`Blocked`** — report to user; do not substitute in-thread execution.
+- **Only exception:** user explicitly overrides harness dispatch for this turn (document the override).
+- Concurrent writers / QC cwd alignment → **`mstar-branch-worktree`** (not a separate “mode”).
+- Implement subagents: recipient is already `Execute as`; **no** recursive Task with same `subagent_type`. Assignment wins (`Delegation: forbidden` unless stated).
 
 ## Clarify
 

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

@@ -36,22 +36,9 @@ Use host subagent / Task / equivalent per **`parallel-dispatch.md`**. OpenCode P
 - Large unrelated platform injections (e.g. long ecosystem prompts): prefer on-demand / `alwaysApply: false` when not stack-relevant.
 - One default channel per capability class (search, docs).
 
-## Optional MCPs / skills (user-enabled)
-
-Aligned with `mstar-harness-core` `references/open-harness-principles.md`. Editing global `opencode.json` requires explicit user consent.
-
-| Capability | Purpose | Notes |
-|------|------|------|
-| **Current docs** | Versioned API/libs | Context7-like MCP or equivalent |
-| **Web search** | Time-sensitive / migrations | Avoid duplicate search MCPs |
-| **Code pattern search** | Cross-repo references | e.g. grep.app MCP |
-| **Repo graph** | Impact / PR risk | e.g. GitNexus |
-| **Browser / E2E** | Observable QA evidence | agent-browser, Playwright |
-| **Git workflow** | Atomic commits, branch closure | git-commit, finishing-a-development-branch |
-| **Systematic debugging** | RCA before fix | `mstar-coding-behavior` |
-| **OpenViking memory** | Long-term memory tools | Only if `memsearch` present; `openviking-memory-plugin.md` |
-
-**Not recommended**: overlapping search MCPs; extra tools to mask missing harness baseline.
+## Standalone harness note
+
+Bundled **`mstar-*` skills** are self-contained in this repository. User-installed host MCPs, external skills, or CLIs are **outside harness SSOT** — do not add them to `mstar-*` load order or treat them as required for gates.
 
 ## Maintenance boundary
 

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

@@ -2,7 +2,7 @@
 
 Shared PM dispatch contract for **any** host that uses subagent / Task / named-role invoke (OpenCode, Cursor Task, Codex only when a callable multi-agent / Task tool is actually available). Process SSOT also in `mstar-dispatch-gates`.
 
-If the active host has no callable invoke tool, this reference does not create delegation capability: use single-session execution, multi-window handoff, or mark dispatch `Blocked`.
+If the active host has no callable invoke tool, this reference does not create delegation capability — mark dispatch **`Blocked`** and report to the user. Do not substitute PM-thread or single-session role execution unless the user explicitly overrides harness dispatch for this turn.
 
 ## Paste-only failure
 

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

@@ -22,10 +22,20 @@ iteration-start → [per-plan lifecycle × N] → iteration-close → PR → mer
      │ Review & Edit chain → PM lock                 │ 标记迭代完成
      │ 创建 integration 分支                          │ commit 到 integration 分支
      │                                               │
-     └──────── iteration-drive（跨 plan 追踪）────────┘
+     └──────── Phase 2（Autonomous Execute）────────┘
 ```
 
-**关键定位**：iteration-close 是 iteration-drive 命令的**最后一个 Phase**，在 integration 分支上执行，**完成后再创建 PR**。所有 compound 产物（knowledge docs、compass 更新、CONCEPTS.md）作为迭代交付的一部分随 PR 合入已记录的 `target_branch`。一次迭代 = 一个 PR。
+**关键定位**：Phase 3 iteration-close 是 **Autonomous Execute 之后的独立 Phase**，在 integration 分支上执行，**完成后再创建 PR**。所有 compound 产物（knowledge docs、compass 更新、CONCEPTS.md）作为迭代交付的一部分随 PR 合入已记录的 `target_branch`。一次迭代 = 一个 PR。
+
+## Phase transition gates（HARD — 防跳步）
+
+| 边界 | 触发 | 必须 | 禁止 |
+|------|------|------|------|
+| **→ Phase 3** | `status.json` 中 compass 登记的全部 plan 均为 `Done` | 打印 `## Phase 3: iteration-close`；执行 §3.0→§3.5；host todo `phase-3-iteration-close` 保持 open 直至 §3.5 | 开 PR；宣称 iteration 完成；仅依赖 final plan closure |
+| **→ PR** | §3.5 exit checklist 全 `[x]`；frontmatter `status: completed` + `end_date` | commit integration 分支 → PR 到 `metadata.target_branch` | 跳过 §3.1 entry checklist 或 compound Phase 6 |
+| **iteration-start → integration branch** | §1.6 Review & Edit chain | 三角色 **按序** invoke 完成 + compass `status: locked` | PM 线程代做三角色编辑；并行三角色；review 前 commit |
+
+**误判信号**：对话里出现 compound 摘要、roadmap 更新、或「所有 plan 已完成」但 **未** 打印 §3.1 / §3.5 checklist → 视为 **Phase 3 未执行**，回到 §3.0。
 
 **per-plan 状态 SSOT**：`{HARNESS_DIR}/status.json`（per-plan Todo/InProgress/InReview/Done）。
 **迭代状态 SSOT**：`{ITERATION_DIR}/<id>-delivery-compass.md` frontmatter `status` + `{ITERATION_DIR}/README.md` 索引。
@@ -64,6 +74,8 @@ PM 在新迭代启动时执行。
 
 **Branch policy gate**：若用户、现有 roadmap、或项目约定未明确 `iteration_base_branch` / `target_branch`，PM 必须检查当前分支并向用户确认。**不得**因为存在 `main` / `master` 就默认从默认分支开 iteration 或向默认分支提 PR。
 
+**Direction lock**：compass/plans 初稿落盘前，PM 须与用户逐问收敛**单一**迭代方向、成功标准、非目标，并确认 delivery branch policy；决策写入 compass `## Scope` / `## Acceptance Criteria` / `## Non-Goals` 与 Delivery Branch Policy。
+
 ### 1.3 创建迭代 compass
 
 写入 `{ITERATION_DIR}/<iteration-id>-delivery-compass.md`。**必须**使用 `references/iteration-compass-template.md` 完整结构（YAML frontmatter + `## Roadmap Position` + close 占位节）。`end_date` 仅在 iteration-close 填入；禁止用正文 completion prose 替代 frontmatter `status`。
@@ -123,7 +135,7 @@ plans: []
 
 ### 1.5 登记到 status.json（formal iteration 必填）
 
-iteration-start / iteration-drive 正式流程**必须**写入 `{HARNESS_DIR}/status.json`：
+iteration 正式全流程**必须**写入 `{HARNESS_DIR}/status.json`：
 
 - root `metadata.iteration_base_branch` — 创建 `spec_integration_branch` 的祖先 ref（**不是**隐式 `main`）
 - root `metadata.target_branch` — iteration-close 后 PR 的目标分支
@@ -133,18 +145,18 @@ compass frontmatter 的 `iteration_base_branch` / `target_branch` **必须与**
 
 ### 1.6 Review & Edit chain（integration 分支前强制）
 
-**iteration-start 在 PM lock 前不算完成**——compass/plans 初稿落盘 ≠ Done。
-
-与 host 命令 `iteration-start` §5 对齐：产物 SSOT 在本节；派发机制 SSOT 在命令 + **`mstar-dispatch-gates`**。
+**Phase 1 在 PM lock 前不算完成**——compass/plans 初稿落盘 ≠ Done。
 
-PM **不得**将迭代 harness 文档 commit 到 `spec_integration_branch`，直到：
+派发机制 → **`mstar-dispatch-gates`**（specialist review-and-edit dispatch，**顺序链**）。PM **不得**将迭代 harness 文档 commit 到 `spec_integration_branch`，直到：
 
-1. **@product-manager**、**@architect**、**@writing-specialist** 已通过宿主 Task（或等价角色 invoke）**直接编辑**（非仅评论）compass、plans 及受影响 specs
+1. **@product-manager** → **@architect** → **@writing-specialist** 已按序通过宿主 invoke **直接编辑**（非仅评论）compass、plans 及受影响 specs；每一环基于上一环落盘修订
 2. PM 将 compass `status` 设为 `locked`，并确认各 plan 的 Prepare gate（specify / clarify / plan）
 
+**顺序理由**：产品范围与优先级 → 架构与契约 → 术语与行文；并行会导致后手重复劳动或覆盖前手未定稿内容。
+
 **完成证据** = 磁盘上的 compass / plans / specs 修订 + compass `status: locked`。**不**要求 `reports/<iteration-id>/` 审查报告——迭代审查的 SSOT 是被编辑的文档本身，无 per-plan QC 式审计链。
 
-**反模式**：PM 线程代替三角色完成全部编辑而不派发 Task —— 见 **`mstar-harness-core`** 反模式索引。
+**反模式**：PM 线程代替三角色完成全部编辑而不 invoke；或将本链三角色并行派发 —— 见 **`mstar-harness-core`** 反模式索引。
 
 ---
 
@@ -158,10 +170,10 @@ PM **不得**将迭代 harness 文档 commit 到 `spec_integration_branch`，直
 
 1. `{HARNESS_DIR}/status.json` 中至少一条 plan `status` ≠ `Done`
 2. **Pre-implement gate = GO**：plan 已 locked、tasks ready（见 `mstar-phase-gates`）
-3. 用户意图为 **continue Execute**（`/pm`、`/iteration-drive`、"推进 iteration" 等）
+3. 用户意图为 **continue Autonomous Execute**（推进迭代 Execute、继续 per-plan 循环等）
 4. **Branch metadata gate**：root `metadata.iteration_base_branch`、`metadata.target_branch` 已登记，且至少一条 active plan 有 `metadata.spec_integration_branch`（或可从 compass 同轮 backfill）。**缺失 → STOP**，不得用 `main`/`master` 补位。
 
-任一 false → **stop**。Prepare 未完成 → 引导先跑 `/iteration-start`。
+任一 false → **stop**。Phase 1 / Prepare 未完成 → 先完成 Phase 1 或 per-plan Prepare，再进入本 Phase。
 
 ### 2.1 Session todos（派发前设护栏）
 
@@ -169,7 +181,7 @@ PM **不得**将迭代 harness 文档 commit 到 `spec_integration_branch`，直
 
 | Host | 工具 | 最小集合 |
 |------|------|---------|
-| **Cursor** | `TodoWrite` / CreatePlan todos | 当前 `plan_id`；下一批 gates（implement/QC/QA）；分支 checkpoint |
+| **Cursor** | `TodoWrite` / CreatePlan todos | 当前 `plan_id`；下一批 gates（implement/QC/QA）；分支 checkpoint；**仅剩 1 个非 Done plan 时追加 `phase-3-iteration-close`**（Done 后保持 open 直至 §3.5） |
 | **Codex** | `update_plan` / Goal UI | 同上 |
 | **OpenCode** | host todo/plan UI（如有） | 同上 |
 
@@ -211,10 +223,20 @@ SSOT = `{HARNESS_DIR}/status.json` + `{PLAN_DIR}/`。todos 只追踪本轮下一
 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
 
-全部 plan `Done` → **STOP** per-plan loop（§2.4 结束），打印 **`## Phase 3: iteration-close`**，按 §3.0 起独立执行。final plan 的 Assignment / closure 可以提供素材，但不能替代 Phase 3 gate。
+全部 plan `Done` → **Phase transition gate**（见上文 **Phase transition gates**）：
+
+1. **STOP** per-plan loop — 禁止 merge 后继续下一 plan、禁止开 PR、禁止会话结束语。
+2. 打印 **`## Phase 3: iteration-close`**。
+3. 按 §3.0 起独立执行至 §3.5。final plan 的 Assignment / closure 仅作输入，**不能**替代 Phase 3 gate。
 
 ### 2.5 Dispatch-first（implement 派发约束）
 
+**派发回合纪律**（Cursor Task / 具名 invoke 宿主）：
+
+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。
+
 | Do | Don't |
 |----|------|
 | **Loop:** `## Assignment` → invoke → Completion Report v2 → 更新 status → next | PM 亲自 Write/Edit/Shell 产品代码 |
@@ -235,7 +257,7 @@ SSOT = `{HARNESS_DIR}/status.json` + `{PLAN_DIR}/`。todos 只追踪本轮下一
 
 ## Phase 3: iteration-close（收口迭代）
 
-PM 在迭代内全部 plan Done 后执行。**本 Phase 在 integration 分支上运行**，产出物 commit 到 integration 分支，随迭代 PR 合入 root `metadata.target_branch`。触发方：`commands/iteration-drive.md`（Autonomous Execute 全部 Done 后进入）。
+PM 在迭代内全部 plan Done 后执行。**本 Phase 在 integration 分支上运行**，产出物 commit 到 integration 分支，随迭代 PR 合入 root `metadata.target_branch`。入口：Phase 2 全部 plan `Done` 后按 **Phase transition gates** 进入。
 
 **Close Done 定义**：§3.1→§3.5 全部完成；compass frontmatter 写入 `status: completed` + `end_date`；每篇新增 knowledge doc 已登记 `{KNOWLEDGE_DIR}/README.md`。只在 final plan 中写了 compound / roadmap / PR 说明，不算 iteration-close 完成。
 
@@ -358,7 +380,7 @@ PR 目标使用 root `metadata.target_branch`；缺失时停止并补齐，不
 ## NOT to do
 
 - 不要在 per-plan Done 后立即单独 compound——等 iteration-close 统一做
-- 不要在 iteration-drive 中修改 per-plan gate 判定
+- 不要在 Autonomous Execute（Phase 2）中修改 per-plan gate 判定
 - 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`

package/harness-skills/mstar-iteration/references/iteration-compass-template.md

@@ -92,25 +92,25 @@ Status values: `Todo` | `InProgress` | `InReview` | `Done` | `Blocked`
 
 | Field | Required | When to fill |
 |-------|----------|-------------|
-| `iteration_id` | Yes | iteration-start |
-| `start_date` | Yes | iteration-start |
-| `end_date` | No | Add at iteration-close §3.4 only |
-| `status` | Yes | iteration-start → `active` / `locked`（§1.6 PM lock）；iteration-close §3.4 → **`completed`**（YAML frontmatter，非 prose completion status） |
-| `iteration_base_branch` | Yes | iteration-start; integration branch must be created from this ref |
-| `target_branch` | Yes | iteration-start; final PR target after iteration-close |
-| `plans` (frontmatter) | Recommended | iteration-start (initial), iteration-drive (add new) |
-| `## Plans` table | Yes | iteration-drive (sync status), iteration-close (final) |
-| `## Milestones` | Recommended | iteration-start, iteration-drive (update) |
-| `## Acceptance Criteria` | Yes | iteration-start |
-| `## Non-Goals` | Yes | iteration-start |
-| `## Roadmap Position` | **Yes** | iteration-start（必填节，非散落于 general context prose）；iteration-close §3.3（current iteration → `delivered`） |
-| `## Risk Register` | Optional | iteration-start, iteration-drive (update) |
-| `## Compound Round Summary` | Yes | iteration-close §3.4 |
-| `## Iteration Retrospective` | Recommended | iteration-close §3.4 |
+| `iteration_id` | Yes | Phase 1 |
+| `start_date` | Yes | Phase 1 |
+| `end_date` | No | Phase 3 §3.4 only |
+| `status` | Yes | Phase 1 → `active` / `locked`（§1.6 PM lock）；Phase 3 §3.4 → **`completed`**（YAML frontmatter，非 prose completion status） |
+| `iteration_base_branch` | Yes | Phase 1; integration branch must be created from this ref |
+| `target_branch` | Yes | Phase 1; final PR target after Phase 3 |
+| `plans` (frontmatter) | Recommended | Phase 1 (initial), Phase 2 (add new) |
+| `## Plans` table | Yes | Phase 2 (sync status), Phase 3 (final) |
+| `## Milestones` | Recommended | Phase 1, Phase 2 (update) |
+| `## Acceptance Criteria` | Yes | Phase 1 |
+| `## Non-Goals` | Yes | Phase 1 |
+| `## Roadmap Position` | **Yes** | Phase 1（必填节，非散落于 general context prose）；Phase 3 §3.3（current iteration → `delivered`） |
+| `## Risk Register` | Optional | Phase 1, Phase 2 (update) |
+| `## Compound Round Summary` | Yes | Phase 3 §3.4 |
+| `## Iteration Retrospective` | Recommended | Phase 3 §3.4 |
 
 ## Legacy compass drift
 
 若仓库中已有 compass **无** YAML frontmatter、**无** `## Roadmap Position`，或只有 prose completion status：
 
-- **新迭代**：iteration-start 必须按本模板重写，不延续 prose-only 惯例。
-- **收口时**：`mstar-iteration` §3.0.5 在 iteration-close 入口规范化，再执行 §3.3–§3.4。
+- **新迭代**：Phase 1 必须按本模板重写，不延续 prose-only 惯例。
+- **收口时**：`mstar-iteration` §3.0.5 在 Phase 3 入口规范化，再执行 §3.3–§3.4。

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

@@ -1,6 +1,6 @@
 ---
 name: mstar-roles
-description: Morning Star role prompt hub — `agents/*.md` shells plus full behavior in `references/*.md`, each with a **Required Skill Dependencies** list (which `mstar-*` topic skills to load after `mstar-harness-core`). Always load for any Morning Star role (`project-manager`, `product-manager`, `architect`, `fullstack-dev`, `fullstack-dev-2`, `frontend-dev`, `qa-engineer`, `qc-specialist*`, `ops-engineer`, `writing-specialist`, `prompt-engineer`). Cross-role summary table in this SKILL.md; per-role lists are authoritative for that role's session.
+description: Morning Star role prompt hub — `agents/*.md` shells plus full behavior in `references/*.md`, each with a **Required Skill Dependencies** list (which `mstar-*` topic skills to load after `mstar-harness-core`). Always load for any Morning Star role (`project-manager`, `product-manager`, `architect`, `fullstack-dev`, `fullstack-dev-2`, `frontend-dev`, `qa-engineer`, `qc-specialist*`, `ops-engineer`, `writing-specialist`, `prompt-engineer`). Cross-role **Role → typical topic skills** summary in this SKILL.md; per-role lists in `references/*.md` are authoritative for that role's session. Full topic skill index → **`mstar-harness-core`**.
 ---
 
 ## Load Order (Required)
@@ -33,28 +33,6 @@ If any conflict appears, `mstar-harness-core` remains the authoritative source f
 | `writing-specialist` | `references/writing-specialist.md` | — |
 | `prompt-engineer` | `references/prompt-engineer.md` | — |
 
-## Shared Skill Dependencies
-
-Treat these as baseline dependencies **where the role touches implementation, review, verification, or ops execution** (see `mstar-harness-core` load contract). Load **on demand**, not as a fixed bundle.
-
-| Skill | Use when task involves |
-| --- | --- |
-| `mstar-harness-core` | **Always** (non-trivial work): entry, state machine, Task category, explore boundary, skill index |
-| `mstar-phase-gates` | Prepare/Execute gates, clarify, hotfix path, intention gate |
-| `mstar-dispatch-gates` | PM dispatch; **all leaf executors** before any Task/subagent call |
-| `mstar-branch-worktree` | Git write, parallel worktrees, QC/QA checkout fields |
-| `mstar-plan-conventions` | `{HARNESS_DIR}` discovery, init, Spec branch naming, plan-writing path gate |
-| `mstar-plan-artifacts` | Main plan, `reports/`, `status.json`, residual, knowledge/iteration, Done compaction |
-| `mstar-design-md` | DESIGN.md design system spec — create/audit/maintain tokens, completeness checklist, light/dark dual-theme |
-| `mstar-review-qc` | QC workflow, template, verdict, high-risk checks |
-| `mstar-coding-behavior` | Implementation/debug/refactor/review-feedback handling (**not** PM orchestration-only) |
-| `mstar-skill-authoring` | New skills, major skill rewrites, trigger descriptions, behavior-shaping skill validation |
-| `mstar-compound` | Capturing solved problems as structured knowledge in `{KNOWLEDGE_DIR}`; Discoverability check; CONCEPTS.md synergy |
-| `mstar-compound-refresh` | Auditing/updating/consolidating knowledge docs; maintaining `{KNOWLEDGE_DIR}` accuracy |
-| `mstar-iteration` | Iteration lifecycle: iteration-start (scope/roadmap lock), Autonomous Execute (per-plan dispatch loop), iteration-close (compound + roadmap update) |
-| `mstar-strategy` | Creating/maintaining `STRATEGY.md`; project vision, principles, technology direction |
-| `mstar-host` | Host-specific behavior (auto-detect; `references/opencode.md` / `cursor.md` / `codex.md`) |
-
 ### Role → typical topic skills (after `mstar-harness-core`)
 
 | Role | Typical adds |

package/harness-skills/mstar-roles/references/project-manager.md

@@ -116,6 +116,9 @@ If any item below matches, fix the dispatch/plan state or mark `Blocked`—do **
 - **NEVER** label `QA: skipped` for report-only QA—still dispatch `@qa-engineer` with report-only mode; QC skip rules are separate and explicit.
 - **NEVER** let non-PM/non-QA roles mark plan `Done`.
 - **NEVER** accept “temporary workaround”, “follow-up later”, “next plan”, or “split into batches” as narrative-only scope management. If work is deferred or staged, write the roadmap/tracking location before implement GO or Done.
+- **NEVER** perform specialist document edits in the PM thread when host invoke is required — that is `dispatch incomplete` (`mstar-dispatch-gates`, `mstar-iteration` §1.6).
+- **NEVER** mark the last plan `Done` and then create a PR or declare the iteration complete without **`## Phase 3: iteration-close`** and `mstar-iteration` §3.1–§3.5 checklists.
+- **NEVER** treat final plan closure prose as iteration-close — compound, roadmap `delivered`, and compass `status: completed` require Phase 3.
 
 ---
 

package/harness-skills/mstar-roles/references/prompt-engineer.md

@@ -51,7 +51,7 @@ If any item below matches, **stop** and return `Blocked` to `project-manager` in
 When creating a new skill, making a major skill rewrite, changing frontmatter `description`, or changing behavior-shaping skill text:
 
 1. MUST read `mstar-skill-authoring` before editing.
-2. MUST follow the repository `skill-creator` requirement when the change is a new skill or major rewrite.
+2. MUST follow the repository **`skill-creator`** requirement when the change is a new skill or major rewrite (`AGENTS.md` → Skill-Creator Requirement).
 3. MUST include validation evidence for behavior-shaping changes, or explicitly state why only manual/search validation is practical.
 
 ## Prompt Change Minimal Checklist

package/harness-skills/pm/SKILL.md

@@ -40,37 +40,9 @@ Prepare/Execute gates, routing, Assignment templates, Task Board, QC tri-review,
 
 **Exceptions:** user explicitly asks PM thread to implement; hotfix per `mstar-phase-gates`.
 
-## iteration-start dispatch
+## Iteration
 
-After grill-me + draft compass/plans（**`mstar-iteration` § Phase 1**）:
-
-- **Do not** commit to `iteration/<iteration-id>` until Review & Edit chain completes（§1.6）。
-- Dispatch **@product-manager**, **@architect**, **@writing-specialist** via host Task — **one invoke per role minimum**（parallel when independent）。
-- PM final lock is PM-whitelist work（compass `status: locked`, merge conflicts, Prepare gate confirmation）。
-- **Not** PM-whitelist: performing all three specialist document edits without subagent Task.
-
-Before §6 Integration Branch, print **`iteration-start` pre-commit checklist**（command §5）; all items must be `[x]`。
-
-## iteration-close（drive / 全部 plan Done 后）
-
-**`mstar-iteration` § Phase 3** 为独立 Phase；final plan closure 只能作为输入。
-
-- 全部 plan `Done` 后 **STOP** loop，打印 Phase 3 入场，从 §3.0 执行至 §3.5。
-- §3.1 close entry checklist 与 §3.5 close exit checklist 必须在对话中打印；全部为 `[x]` 方可 commit / PR。
-- 每篇新增 compound doc 必须完成 `mstar-compound` Phase 6（`{KNOWLEDGE_DIR}/README.md`）。
-- compass 完成形式：frontmatter `status: completed` + `end_date` + `## Roadmap Position` current iteration `delivered`——prose completion status 不算。
-
-## iteration branch policy（start + drive）
-
-| 字段 | 含义 | 禁止 |
-|------|------|------|
-| `iteration_base_branch` | 创建 `spec_integration_branch` 的祖先 ref | 未记录时假定 `main` |
-| `spec_integration_branch` | plan feature 分支 merge 靶 | 从当前 checkout 随意开枝 |
-| `target_branch` | iteration-close 后 PR 目标 | 未记录时 PR 到 `main` |
-
-- **iteration-start**：grill-me / §6 与用户锁定；写入 compass frontmatter + `status.json` `metadata`。
-- **iteration-drive**：§2.0 #4 + §2.3 执行前校验；创建 integration 分支必须用记录的 base。
-- QC `Review range` merge-base：用 `target_branch` 或 PM 书面 ref（`mstar-iteration` §2.3），禁止无依据写死 `origin/main`。
+Formal iteration（Phase 1 → Phase 2 → Phase 3）→ **`mstar-iteration`** only. Do not duplicate phase gates, branch policy, or close checklists here.
 
 ## Cursor Plan mode
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mstar-harness/opencode",
-  "version": "0.7.5",
+  "version": "0.7.7",
   "description": "Morning Star harness OpenCode plugin (skills bootstrap and agent loading).",
   "license": "MIT",
   "repository": {

package/harness-skills/mstar-harness-core/references/library-docs-protocol.md

@@ -1,35 +0,0 @@
-# Context7 Library Docs Protocol (Shared)
-
-This protocol is shared by all hosts in Morning Star. Host adapter skills should consume this section, not redefine it.
-
-## When to use
-
-- The question is about library/framework/SDK/CLI/cloud API syntax, config options, migration behavior, or version-specific usage.
-- The answer quality depends on current docs rather than model memory.
-
-## When not to use
-
-- Pure refactoring strategy.
-- Writing scripts from scratch without external API dependence.
-- Business logic debugging not tied to third-party docs.
-- Generic programming concepts.
-- PR/code review judgment.
-
-## Single-path policy
-
-Use one path per question. Do not run MCP and CLI in parallel for the same library/question pair.
-
-1. **Primary path: Context7 MCP (if available)**
-   - Read tool schema first.
-   - Query with the user's full question.
-   - Prefer versioned library IDs when version is known.
-2. **Fallback path: ctx7 CLI (only if MCP unavailable/fails)**
-   - `npx ctx7@latest library <name> "<user question>"`
-   - `npx ctx7@latest docs <libraryId> "<user question>"`
-   - Max 3 commands per question.
-
-## Guardrails
-
-- Never include secrets in commands or queries.
-- Switch to fallback only after primary path fails.
-- Report uncertainty when docs retrieval fails on both paths.

package/harness-skills/mstar-harness-core/references/open-harness-principles.md

@@ -1,76 +0,0 @@
-# 开源 Agent Harness 理念（Morning Star）
-
-## 定位
-
-本文归纳从公开 agent harness 实践中提炼的**流程理念**，并已写入 `mstar-harness-core`、`mstar-roles` skill 的 `project-manager` 角色 等，作为本配置的默认行为约定。
-
-## 已内化的原则（执行时必须遵守）
-
-| 理念 | 含义 | 在本仓库的落点 |
-|------|------|----------------|
-| **意图优先于字面** | 先弄清「用户真正要达成什么」，再分类与分派 | `mstar-phase-gates` SKILL.md Prepare · `clarify` / `plan` 意图门禁；PM「第一性原理」 |
-| **先准备再实现** | 访谈式规划、锁范围、再写代码 | Prepare：`specify -> clarify -> plan`；Execute：`plan locked -> tasks -> implement` |
-| **`clarify` 核心纪律** | 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；每问带推荐答案 | **`mstar-phase-gates`** SKILL.md（含 Phase Gate Playbook） |
-| **按任务类别选能力与模型** | 视觉/深读/快改/硬逻辑等用不同强项 | `mstar-harness-core` SKILL.md「Task category」；Assignment 字段 **`Task category`**；宿主侧按角色配置 model（如 OpenCode 的 `opencode.json`） |
-| **可验证编辑** | 减少「凭记忆 Patch」导致的漂移与损坏 | `mstar-phase-gates` SKILL.md「可验证编辑与上下文纪律」：读后再改、失败则重读 |
-| **持续推进与可核对完成** | 长任务有清单、有关门证据，避免空转 | `mstar-phase-gates` / `mstar-review-qc`；实现侧自检见 `mstar-coding-behavior` |
-| **编码行为约束（轻量）** | 降低静默假设、过度设计与无关改动 | `mstar-coding-behavior`：Think Before Coding / Simplicity First / Surgical Changes / Goal-Driven Execution |
-| **并行与边界** | 多线任务不踩同一写归属、不绕过分支门禁；**开发**阶段同仓多可写并发须独立 **`git worktree`**；**QC / QA** 在 **同一检出**（`Review cwd`）与 **同一 `plan_id` + `Review range` / `Diff basis`** 上审查与验证，保证三票同一功能 | `mstar-branch-worktree`；`mstar-review-qc`；`mstar-dispatch-gates` |
-| **分层上下文（可选）** | 大仓库用目录级 `AGENTS.md` 降噪；根 `AGENTS.md` 维护边界见下文专节 | `mstar-harness-core` SKILL.md「分层上下文」；由业务项目维护者按需添加 |
-| **结构化澄清（按宿主）** | 向用户澄清/抉择时，**有 `question` 类能力则优先**；否则结构化正文；长问兜底 | `mstar-phase-gates` SKILL.md Prepare · `clarify`；当前宿主的 `mstar-host` skill；`mstar-roles` · `project-manager` |
-
-## 与常见 harness 说法的对照（帮助理解角色分工）
-
-| 常见说法 | 本仓库实体 |
-|----------|------------|
-| 总编排 | `@project-manager` |
-| 规划/访谈 | `@product-manager` / `@architect` + Prepare 阶段 |
-| category 路由 | **`Task category`** + 路由表 + 子代理选择 |
-| 持续推进 / 不半途而废 | Phase Gate + Todo/`tasks` + 验证门禁 |
-| 行级哈希锚定编辑 | 本仓库以 **读后再改 + 小步 Patch** 纪律落实（见 `mstar-harness-core`） |
-
-## 项目根 `AGENTS.md` 写什么、不写什么（建议）
-
-业务仓库根目录（或分层中的目录级）`AGENTS.md` 适合承载：**长期有效的规则、技术栈约束、工作流门禁**；不宜承载易变状态或已另有 SSOT 的内容，以免双轨漂移。
-
-### 宜放入
-
-- 技术栈与架构边界（变更时更新）
-- 开发与验证命令、目录布局、命名约定
-- 与 harness 对齐的**持久**规则（分支策略引用、**`{HARNESS_DIR}`** / **`{PLAN_DIR}`** 路径说明等）
-
-### 不宜放入（应指向专门落点）
-
-| 内容 | 更合适的落点 |
-|------|----------------|
-| 详细计划状态、阻塞列表、当前 sprint 叙述 | `{HARNESS_DIR}/status.json`（及可选 `{HARNESS_DIR}/notes.json` 程序时间线） |
-| 规格正文、枚举定义、API 契约全文 | 项目约定的冻结规格目录（如 `docs/spec/`、`.agents/designs/...`，名称自定） |
-| 迭代/版本级 delivery compass、program 范围快照 | `{ITERATION_DIR}`（并维护索引 README，见 `mstar-plan-conventions`） |
-| 某一 plan 的评审稿、gap 分析、跨版本实施 SSOT | `{KNOWLEDGE_DIR}`（并维护索引 README，见 `mstar-plan-conventions`） |
-| 临时 workaround、仅本轮有效的结论 | 主 plan 文件或 knowledge，收口后提炼再考虑进 `AGENTS.md` |
-
-### 更新前自检（避免污染单一权威）
-
-- 是否为**持久规则**，而非一两次性的当前状态？
-- 是否适用于**整个项目**，而非单个 plan？
-- 是否与冻结规格**重复**？若重复，应改为「引用路径」而非粘贴全文。
-- 是否其实属于 **`{HARNESS_DIR}/status.json`**（状态、门禁摘要、时间线）？
-
-### 典型信息源层级（模板）
-
-项目可自定义目录名；逻辑顺序常为：
-
-1. **冻结规格 / 设计权威**（版本化、显式变更流程）
-2. **根 `AGENTS.md`** — 项目级规则与约定
-3. **`{HARNESS_DIR}/status.json`** — 当前执行状态与 open residual 的 SSOT
-4. **`{ITERATION_DIR}`** — 迭代/版本级 compass（可选）
-5. **`{KNOWLEDGE_DIR}`** — 支撑实施的上下文文档（可选）
-
-若 `AGENTS.md` 与冻结规格冲突，**以规格为准**，并修订 `AGENTS.md` 对齐。细节分工与可到达性要求见 `mstar-plan-conventions`。
-
-## 延伸阅读
-
-- 按能力选配 MCP/skills：OpenCode → `mstar-host` `references/opencode.md`；其它宿主见 `mstar-host` 检测表
-- Evidence and checkpoints: `mstar-phase-gates` / `mstar-review-qc`; implementation self-checks: `mstar-coding-behavior`
-- 库文档检索共享协议：`mstar-harness-core` `references/library-docs-protocol.md`
-- 跨角色编码行为准则（轻量、可复用）：`mstar-coding-behavior`

package/harness-skills/mstar-harness-core/references/openviking-memory-plugin.md

@@ -1,45 +0,0 @@
-# OpenViking Memory Plugin (OpenCode, optional)
-
-This reference applies **only** when the current agent session exposes OpenViking tools (detection: **`memsearch`** is available in the tool list). It documents harness-aligned usage; implementation details follow the OpenCode plugin (for example `openviking-memory.ts` beside `openviking-config.json` under the user’s OpenCode plugins directory).
-
-## What the plugin provides
-
-Typical tools (names match the reference plugin):
-
-| Tool | Purpose |
-| --- | --- |
-| `memsearch` | Unified semantic search over memories, resources, and skills (`mode`: `auto` / `fast` / `deep`). |
-| `memread` | Read a single item by `viking://` URI with progressive depth (`abstract` / `overview` / `read` / `auto`). |
-| `membrowse` | List / tree / stat views under a `viking://` prefix. |
-| `memcommit` | Trigger session commit / memory extraction for the mapped OpenViking session (optional mid-session). |
-
-**URI rule**: `memread` and `membrowse` require URIs starting with `viking://` (validated by the plugin).
-
-**Service dependency**: The plugin calls an OpenViking HTTP API (default `http://localhost:1933` unless overridden). If tools error with connection or health failures, treat memory as **unavailable** for this turn; do not block core harness gates on memory.
-
-**Auto features** (when enabled in plugin config): conversation capture, periodic auto-commit, and optional **auto-recall** injection (`<relevant-memories>` appended to the latest user message). Recall is best-effort and must not replace explicit search when you need traceable evidence.
-
-## Harness alignment (mandatory)
-
-1. **Subordinate to SSOT**: `mstar-harness-core` state machine, phase gates, branch/worktree rules, and QC/QA alignment **always** override any suggestion retrieved from memory. If memory conflicts with plan, `status.json`, or Assignment, **follow the written artifacts** and record the conflict in notes or Completion Report.
-
-2. **No secrets in memory tools**: Do not paste API keys, tokens, or private credentials into `memsearch` queries or stored memories. Redact before commit-style operations.
-
-3. **Evidence for claims**: Memory hits are **hints**, not proof. For library/API facts, still follow `references/library-docs-protocol.md` (Context7 MCP / ctx7) when the question depends on current docs.
-
-4. **When to call `memsearch`**: Prefer early in a **new** task or thread when user preferences, prior decisions, or plan IDs may exist in OpenViking; after major clarify/plan changes, a fresh search can reduce stale context.
-
-5. **`memread` after `memsearch`**: Use URIs from search results; escalate depth (`overview` → `read`) only when needed to avoid token burn.
-
-6. **`memcommit`**: Use for explicit “persist now” or mid-session extraction when the user asks or when wrapping a milestone. Do **not** spam commits after every trivial edit; the plugin may already run **auto-commit** on an interval—respect that and user policy.
-
-7. **Parallel / multi-agent**: Memory tools do **not** replace PM dispatch, worktree isolation, or QC tri-review invokes. They do not authorize subagent recursion.
-
-## When **not** to rely on this reference
-
-- `memsearch` (and sibling tools) are **absent** → skip this file; no OpenViking rules apply.
-- Non-OpenCode hosts (unless they expose the same tool names with the same semantics) → ignore.
-
-## Configuration (user-owned)
-
-Plugin reads `openviking-config.json` next to the plugin file and env vars such as `OPENVIKING_API_KEY`, `OPENVIKING_ACCOUNT`, `OPENVIKING_USER`. Agents must **not** edit user global config without explicit user consent (see `mstar-harness-core` guardrails).