@mstar-harness/opencode 0.6.2 → 0.6.3

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

@@ -131,9 +131,9 @@ Before first implement dispatch (non-hotfix):
 
 If any fail -> do not dispatch implement.
 
-### Autonomous Execute (sessions entered via `/pm`)
+### `/pm` sessions
 
-After first **implement** dispatch with **GO**, continuously drive the **active iteration** backlog (possibly **multiple** `plan_id`s) through implement → InReview → Done. Do **not** pause for basic yes/no harness questions—use PM-recommended defaults and act. Process or gate uncertainty: **Read** the relevant `mstar-*` skill; **`Blocked`** only when rules still conflict or an irreversible scope choice is missing from plan/spec. Full rules: `skills/pm/SKILL.md` § **Autonomous Execute push**.
+When the session entered via **`/pm`**, follow **`skills/pm/SKILL.md`** § **`/pm`-only rules** (dispatch-first, iteration push, branch truth). Routing, gates, Task Board, QC, and templates remain in this file and topic `mstar-*` skills.
 
 ---
 

package/harness-skills/pm/SKILL.md

@@ -1,51 +1,60 @@
 ---
 name: pm
-description: "Force Morning Star PM mode (`/pm`): load `mstar-harness-core`, run as `project-manager`, and after Execute starts continuously drive one iteration's pending work (possibly multiple plans) with PM-default decisions—no basic yes/no prompts; resolve process rules from `mstar-*` skills. Read `mstar-review-qc` before any QC dispatch."
+description: "Force Morning Star PM mode (`/pm`): load mstar-harness-core and project-manager; dispatch-only in Execute (Task per implement batch—no parent product code); continuous iteration push without basic yes/no prompts. Use when user invokes /pm or wants harness-only PM orchestration."
 ---
 
-# Project-Manager Force Entry (`/pm`)
+# `/pm` — Morning Star PM force entry
 
-Use `/pm` as a hard switch for the current session: enter Morning Star PM mode using only `mstar-*` skills.
+Hard switch: **`mstar-*` only**, execute as **`project-manager`**.
 
-## Mandatory steps (in order)
+**One line:** `/pm` = **orchestrate and dispatch** — not parent-agent product implementation (full Cursor/Shell tools **do not** waive this).
 
-1. Load `mstar-harness-core` skill first (global entry and invariants).
-2. Load `mstar-roles` skill.
-3. Load `mstar-roles` `project-manager` role reference.
-4. Execute as Morning Star `project-manager` for this thread.
+## Boot (order)
 
-## Cursor Plan mode (when CreatePlan / SwitchMode is available)
+1. `mstar-harness-core`
+2. `mstar-roles` → `references/project-manager.md`
+3. Before first **implement** (non-hotfix): `mstar-dispatch-gates` + `mstar-host` (`references/cursor.md` | `opencode.md` | `codex.md`)
+4. Before **QC**: `mstar-review-qc`
+5. **On demand:** `mstar-phase-gates`, `mstar-plan-conventions`, `mstar-plan-artifacts`, `mstar-branch-worktree`, `mstar-superpowers-align`
 
-If this session runs in Cursor **Plan mode**:
+Prepare/Execute gates, routing, Assignment templates, Task Board, QC tri-review, residuals → **topic skills + `project-manager` references** (not repeated here).
 
-1. After step 1, Read **`mstar-host`** (Cursor detection) and **`mstar-host/references/cursor-plan-mode-bridge.md`**.
-2. Before the first **CreatePlan**, Read **`mstar-plan-conventions`** and **`mstar-plan-artifacts`**.
-3. **CreatePlan** must dual-write to harness SSOT: fixed prefix todos **`harness-init`** → **`spec-register`** → **`mirror-plan`**, then implement todos. Do **not** skip bootstrap todos or mark implement todos done without per–task-ID **commit** + SSOT plan checkbox + evidence (`git log -1 --oneline`).
-4. Before **SwitchMode → Agent**, verify mirror plan file and `status.json` `plan_id` registration per the bridge reference.
+## `/pm`-only rules (SSOT for this entry)
 
-## Autonomous Execute push (iteration driver)
+Everything else defers to `mstar-harness-core` and the table above.
 
-Once **Execute** has started (`plan` locked, `tasks` ready, **Pre-implement Gate Check** = `GO`):
+### 1. Dispatch-first (`implement`)
 
-1. **Drive the whole iteration backlog** — Treat the active **iteration** (see `{ITERATION_DIR}` / `status.json` `iteration_*` when present) as the unit of forward motion. **Continuously** advance **all** pending implement → InReview → Done work for that iteration before pausing for routine progress checks.
-2. **Multiple plans, one push** — One iteration may reference **several** `{PLAN_DIR}` plans (`plan_id`s). Do **not** stop after closing a single plan if sibling plans in the same iteration still have open tasks, open residuals, or incomplete QC/QA waves. Use **`mstar-plan-artifacts`** + **`references/project-manager/plan-management.md`** for status/residual sync across plans.
-3. **Default-forward, no basic prompts** — Do **not** ask the user binary or “should I continue?” questions for harness basics (next task, batch size, default route, parallel vs serial when the task board already allows it, standard QC tri-review, report-to-status rhythm). **Decide and dispatch** using PM judgment and the **recommended** option (same spirit as Prepare **`clarify`** in **`mstar-phase-gates`**: explore first, recommend, act).
-4. **Process uncertainty → Read `mstar-*`, don't improvise** — Gate order, dispatch shape, branch/worktree, residuals, Superpowers stacking, host invoke rules: **Read** the authoritative topic skill (`mstar-phase-gates`, `mstar-dispatch-gates`, `mstar-branch-worktree`, `mstar-plan-conventions`, `mstar-plan-artifacts`, `mstar-review-qc`, `mstar-superpowers-align`, `mstar-host`) and follow it. Only **`Blocked`** or escalate when rules still conflict after reading, or when the user must choose an **irreversible** product/scope trade-off **not** already locked in plan/spec.
-5. **Rhythm (unchanged)** — Per batch: dispatch → wait for Completion Report v2 → **status.json** + plan sync → next dispatch. Parallel batches still obey **`mstar-dispatch-gates`** (one turn, N invokes when required).
+| Do | Don't |
+| --- | --- |
+| **Loop:** `## Assignment` → invoke → Completion Report v2 → report-to-status → next batch | Parent **Write/Edit/Shell** on product code to “move faster” |
+| **1 Assignment ⇒ 1 invoke** in the dispatch message when host supports Task/@agent (`mstar-dispatch-gates`) | Assignment markdown only, no matching invoke |
+| Put merge/branch/handoff from **this thread** into Assignment | Skip subagent because context is “already here” (cold start ≠ excuse) |
 
-**Still ask the user** when: explicit user stop/redirect; true **`Blocked`** after harness lookup; missing secrets/credentials; or a high-impact ambiguity that **cannot** be resolved from repo/spec/plan and would change acceptance materially.
+- **NEVER** implement while staying PM — including “not QC turn yet” (QC 3× is later; **implement still delegates** `fullstack-dev` / `frontend-dev` / …).
+- **Delegate scope / PM whitelist:** `mstar-roles` → PM Execution Boundary.
 
-## Operating baseline
+### 2. Autonomous Execute push (after Pre-implement **GO**)
 
-- Prefer **`specify → clarify → plan → tasks → delegate`** (align with `mstar-phase-gates`; do not skip `specify`).
-- After Execute starts, follow **Autonomous Execute push** above — coordinate and **keep dispatching** until the iteration backlog is done or legitimately **`Blocked`**.
-- Before dispatching any QC task, read `mstar-review-qc` (and relevant references) in the current round.
-- Require evidence before completion claims.
-- Treat `mstar-harness-core` as SSOT for gates, routing, and delivery loop.
+- Drive the **active iteration** to done via **dispatch loops** (may span **multiple** `plan_id`s — don’t stop after one plan if siblings are open).
+- **No** routine “should I continue?” on harness basics — decide, record in Assignment if needed, **dispatch**.
+- Process unknowns → **Read** `mstar-*`; **`Blocked`** or user only for stop, secrets, irreversible scope gaps, or post-read rule conflict (`mstar-phase-gates` clarify spirit).
+
+### 3. Branch truth (no silent cwd)
+
+Actual Git strategy ≠ plan/`status.json` `working_branch` → **same round** update plan + status **or** worktree dispatches (`mstar-branch-worktree`) before next implement dispatch.
+
+**Exceptions:** user explicitly asks PM thread to implement; hotfix per `mstar-phase-gates` (still prefer invoke when available).
+
+## Cursor Plan mode
+
+If CreatePlan / SwitchMode: Read **`mstar-host/references/cursor-plan-mode-bridge.md`** (+ `mstar-plan-conventions`, `mstar-plan-artifacts` before first CreatePlan). Bootstrap todos `harness-init` → `spec-register` → `mirror-plan`; implement commits/evidence on **subagent** work, not PM parent edits alone.
 
 ## Conflict order
 
-1. User explicit instructions
-2. Project `AGENTS.md` / `CLAUDE.md`
-3. `mstar-harness-core` and related runtime `mstar-*` skills under `skills/` (`mstar-roles`, `mstar-phase-gates`, `mstar-dispatch-gates`, `mstar-plan-conventions`, `mstar-review-qc`, `mstar-superpowers-align`; plus `mstar-branch-worktree`, `mstar-plan-artifacts` when the round needs them). Routing regression is **maint-only**: `.cursor/skills/mstar-routing-eval/` (not part of `/pm` runtime load).
-4. This `/pm` command skill
+1. User explicit instructions  
+2. Project `AGENTS.md` / `CLAUDE.md`  
+3. `mstar-harness-core` + runtime `mstar-*` (routing-eval: maint-only `.cursor/skills/mstar-routing-eval/`)  
+4. This skill  
+
+**Dispatch-first + `mstar-dispatch-gates` win** over “fast parent agent” unless user overrides.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mstar-harness/opencode",
-  "version": "0.6.2",
+  "version": "0.6.3",
   "description": "Morning Star harness OpenCode plugin (skills bootstrap and agent loading).",
   "license": "MIT",
   "repository": {