Merge pull request #237 from PsychQuant/idd/cluster-idd-verify-engine

cluster: idd-verify engine — DA sequenced-spawn + pai install-time dependency (fork deletion)

Author: kiki830621

plugins/issue-driven-dev/.claude-plugin/plugin.json

@@ -189,6 +189,7 @@ IDD 遵循「**深度整合 >> hard-coded**」原則（#209/#214）：生態系
 
 | Package | 綁定形狀 | 角色 | 缺席行為 |
 | ------- | ------- | ---- | ------- |
+| `parallel-ai-agents`（`psychquant-claude-plugins`） | **Hard dependency** — `plugin.json` `dependencies` 安裝時自動安裝（#219；cross-marketplace，root marketplace 需 `allowCrossMarketplaceDependenciesOn` 增列） | `idd-verify` 的 canonical ensemble 引擎（4 lens + DA + Codex；≥2.18.0 STABLE contract） | 缺席/過舊 → 印一步安裝指令 + fall back **manual fan-out**（品質等價、較慢）；vendored fork 已刪（#207 Residue 成熟） |
 | `superpowers`（`claude-plugins-official`） | **Hard dependency** — `plugin.json` `dependencies` 安裝時自動安裝（#209） | TDD 執行（`idd-implement`）、系統性除錯（`idd-diagnose` bug RCA）、完成前驗證的 canonical process 紀律 | Delegation 點 **fail-fast abort** + 一步安裝指令；絕不 silent degrade |
 | `parallel-ai-agents`（pai） | **Canonical engine + 版本閘門 ≥ 2.18.0**（#207，STABLE external-consumer contract） | `idd-verify` 的 6-AI ensemble 引擎（4 lenses + Devil's Advocate + Codex 編排） | 三層 graceful degrade：canonical → frozen vendored fork → manual fan-out（每層印 notice） |
 | `ralph-loop`（`claude-plugins-official`） | Optional mode 依賴（#28） | `idd-verify --loop` / `idd-all (PR, unattended)` 的 verify-fix loop driver | `--loop` fail-fast + 安裝指引；`idd-all` graceful degrade 到 (direct-commit, attended) |

plugins/issue-driven-dev/README.md

@@ -10,7 +10,7 @@ This discipline has **two halves with very different maturity**, and conflating
 
 | Half | Status | Mechanism |
 |------|--------|-----------|
-| **Read-only diagnose fan-out** | ✅ **real today** | the Workflow tool (deterministic parallel agents + synthesis, exactly like `idd-verify`'s `ensemble-workflow.js`) |
+| **Read-only diagnose fan-out** | ✅ **real today** | the Workflow tool (deterministic parallel agents + synthesis, exactly like `idd-verify`'s canonical pai-ensemble engine（#219 後 vendored fork 已刪，引擎 = parallel-ai-agents plugin）) |
 | **Concurrent *stateful* lanes** (run N full `idd-all` pipelines at once) | ⛔ **deferred** | no primitive exists — within-window agent teams is `## Deferred: Case A` in [`worktree-isolation.md`](worktree-isolation.md) (#167); `TeamCreate` was abandoned by `idd-verify` after #47/#52; a Workflow `agent()` is a single subagent turn that cannot host `idd-all`'s skill orchestration or nest the verify ensemble |
 
 So **`idd-all #a #b #c` batch mode is sequential, not concurrent.** The conflict-class taxonomy below is a **forward-looking safety contract**: it tells you what is safe to parallelize *when you do so manually* (separate Claude sessions / worktrees) or *when a real concurrent-lane primitive eventually lands*. It does NOT assert that any skill auto-parallelizes stateful work today. Do not write a spec or skill that `SHALL`s a concurrency engine that does not exist.