team-narrative

Donchitos ｜ 遊戲開發 ｜ 原文 ｜ MIT

統籌敘事團隊，產出劇情、世界背景設定、對白與關卡敘事

---
name: team-narrative
description: "統籌敘事團隊（narrative-director、writer、world-builder、level-designer），處理劇情、世界背景設定與敘事導向的關卡。"
argument-hint: "[narrative content description] [--review full|lean|solo]"
user-invocable: true
allowed-tools: Read, Glob, Grep, Write, Edit, Agent, AskUserQuestion, TaskCreate, TaskGet, TaskList, TaskUpdate, Bash(bash "*/.claude/skills/team-narrative/../../hooks/yaml-helper.sh" resolve_config *)
model: sonnet
---
# Team Narrative

如果沒有帶參數，就輸出使用說明並結束，不要 spawn 任何 agent：
> 用法：`/team-narrative [narrative content description] [--review full|lean|solo]` — 描述要處理的劇情內容、場景或敘事區域（例如 `boss encounter cutscene`、`faction intro dialogue`、`tutorial narrative`）。這裡不要用 `AskUserQuestion`，直接輸出說明。

帶了參數時，就用結構化的 pipeline 來統籌敘事團隊。

**決策點：** 每個 phase 轉換時，用 `AskUserQuestion` 把 subagent 的提案做成可選項目給使用者選。先在對話中寫出 agent 的完整分析，再用簡短的標籤收下決定。
在 `collaborative` 模式下，使用者核准後才能進下一個 phase。
在 `guided` 模式下，pipeline 會自動往下走，除非某個 phase BLOCKED；
在 `autonomous` 模式下會從頭跑到尾，每個 phase 的結果用 `log_decision` 記錄下來。
屬於 `automation_always_ask` 類別的決策（`is_always_ask_category` helper）不管什麼模式一律要問。參見 `.claude/docs/automation-modes.md`。

## Phase 0: Resolve Config

!`bash "${CLAUDE_SKILL_DIR}/../../hooks/yaml-helper.sh" resolve_config --keys review_mode,automation,team.size`

上面已經解析好了，直接用；`--review` 會覆蓋 `review_mode`。沒有這個區塊的話，就用 `.claude/docs/config-resolution.md` 裡的預設值。

`review_mode` 決定 director gate 的深度，而這條 pipeline 沒有 director gate：
下面任何 phase 在任何 `review_mode` 下，都不會 spawn CD-、TD-、PR- 或 AD-PHASE-GATE。它的 phase gate 就是 pipeline 自己的決策點（定義在下面的 `team.size` 底下），在這些點上工作的 agent 是團隊成員，不是 director gate。

Phase 4 的 ND-CONSISTENCY 在每個模式下都是例外：它是這條 pipeline 自己的審查，不是選用的 director gate，所以 `review_mode` 永遠不會跳過它（誰來跑它還是由 `team.size` 決定，見 Phase 4）。

`automation` 決定上面「決策點」那段的行為。各模式如何改變 pipeline 的行為，見上面的決策點說明和 `.claude/docs/automation-modes.md`。

**`team.size`**：哪些 agent 啟用（跟 review_mode 的 gate 深度和 workflow docs 互相獨立）。
- **`individual`**（預設）：只有 `writer`；`narrative-director` 只在明確出現支柱衝突時才叫。其他 agent 透過 writer 諮詢，不會另外 spawn。
- **`small`**：`narrative-director` + `writer`，加上 `review_mode` 為 `full` 時的 `localization-lead` 和 `world-builder`（在 `lean` 和 `solo` 下，它們透過 writer 諮詢）。
- **`studio`**：Team Composition 裡全部六個 agent，不管 `review_mode` 是什麼（也就是文件中寫的完整 pipeline）。
在 `individual` 下需要用到的非核心 agent，會改由最接近的核心 agent 處理，並附一則說明。**「Phase gate」指的是這條 pipeline 自己列出、以 `AskUserQuestion` 決策點收尾的任何 phase**：可能是上面「決策點」底下的某個轉換，也可能是下面 pipeline 裡寫明的 **Gate** 步驟，**不管 `automation` 是哪個模式都一樣。** `guided` 和 `autonomous` 改變的是 gate 怎麼通過（自動往下走，或用 `log_decision` 記錄），不是它算不算 gate，所以下面的有界例外條件 (3) 在每個模式下的 gate 都成立。被限制成「只在 phase gate 出場」的 agent，只會在這些時點 spawn，其他時候不會。這個「依啟用名單縮小範圍」的規則適用於下面整條 pipeline：任何 phase 只要點名了啟用名單以外的 agent，就改由最接近的核心 agent 處理，不 spawn 它。

**在 Phase 1 之前先公布啟用名單，絕對不准讓縮減悄悄發生。**
spawn 任何東西之前，先用一行講清楚這次實際會 spawn 哪些 agent，以及下面 pipeline 有點名、但在目前 `team.size` 下**不會** spawn 的有哪些。例如：

> `Active set (team.size: <resolved>): <the agents listed for that size above>.`
> `Not spawned this run: <every other agent this pipeline names> — consulted`
> `through <nearest active core agent>. Raise team.size (or modes.rigor) to widen.`

內容要直接從上面的 `team.size` 清單和這個檔案自己的 pipeline 所點名的 agent 填，不要照抄範例。每個 orchestrator 這兩組名單都不一樣。

下面這條 pipeline 看起來像多 agent 的 fan-out，但在預設值下其實只有一兩個 agent：`team-release` 點名了十個、只跑一個，`team-narrative` 在五個 phase 裡點名了六個、只跑 `writer` 一個。**這樣縮減是對的**：`team.size` 是以嚴謹度為前提的，預設收窄就是用來省 token 的。
如果不明講，讀的人就分不出「正確縮減過的 run」和「壞掉的 pipeline」，而上面那條「由最接近的核心 agent 處理並附說明」的規則是在路由當下才觸發，從來不會交代整個 run 的樣子。

這跟這個檔案其他地方的「略過的檢查要回報」是同一條規則：**一個有被執行、卻從沒被講出來的限制，對讀輸出的人來說，跟從來沒執行過的限制看起來一模一樣。**

## Team Composition
- **narrative-director** — 故事弧線、角色設計、對白策略、敘事願景
- **writer** — 撰寫對白、背景設定條目、道具說明、遊戲內文字
- **world-builder** — 世界規則、陣營設計、歷史、地理、環境敘事
- **art-director** — 角色視覺設計、環境視覺敘事、過場／電影化演出的基調
- **level-designer** — 為敘事服務的關卡配置、節奏、環境敘事節拍
- **localization-lead** — 在地化就緒度：標出無法在地化的字串、文化假設與 i18n 缺口

## How to Delegate

用 `Agent` 工具把每個團隊成員 spawn 成 subagent：
- `subagent_type: narrative-director` — 故事弧線、角色設計、敘事願景
- `subagent_type: writer` — 撰寫對白、背景設定條目、遊戲內文字
- `subagent_type: world-builder` — 世界規則、陣營設計、歷史、地理
- `subagent_type: art-director` — 角色視覺設定、環境視覺敘事、電影化基調
- `subagent_type: level-designer` — 為敘事服務的關卡配置、節奏
- `subagent_type: localization-lead` — 在地化就緒度：標出無法在地化的字串、文化假設與 i18n 缺口

**先讀 canon。** 在 Phase 1 之前，先讀現有的共用輸入：敘事聖經或基調指南、`design/narrative/` 底下的角色與陣營檔案，以及世界規則。絕對不要叫 agent 自己發明 canon：如果根本沒有，就講明這一點，並在開始起草之前先問使用者。

**給每個 agent 的 brief 要精簡，不要整包 context 倒過去。** 共用的輸入只讀**一次**，再把整理過的 brief 直接寫在 prompt 裡：只放每個 agent 真正需要的那幾行，你已經讀過的文件絕對不要只丟檔案路徑（agent 拿到路徑會把整份檔案重讀一遍）。只有你還沒讀、而且只有那個 agent 需要的文件，才給路徑。

**每個 agent 的 prompt 結尾都要加上回傳約定：**「Write your full output to `[path]` — that named path is your write authorisation under the bounded exception below, so write it without a separate approval prompt. Return **only** (1) the path written, (2) a ≤5-bullet summary of decisions, (3) any BLOCKED/CONCERNS items, one line each. Do not restate the documents you read.」沒有這個約定的話，agent 會把它讀過的東西全部倒回這個 session。

**`[path]` 要換成真實路徑。** Agent 把草稿寫在
`production/narrative/[content-slug]/` 底下，這在下面的有界例外之內。完成的文件屬於 `design/narrative/`，而例外**不**涵蓋那裡，所以那些由**你**在一次核准之後去寫（見「Write the finals」）：

| Agent | Drafts to (`[path]`) | Final document (you write it) |
|---|---|---|
| narrative-director | `production/narrative/[content-slug]/brief.md` | `design/narrative/[content-slug]-brief.md` |
| world-builder | `production/narrative/[content-slug]/lore-[topic].md` | `design/narrative/lore/[topic].md` |
| writer | `production/narrative/[content-slug]/dialogue-[scene-slug].md` | `design/narrative/dialogue/[scene-slug].md` |
| art-director | `production/narrative/[content-slug]/visual-direction.md` | `design/narrative/[content-slug]-visual-direction.md` |
| level-designer | `production/narrative/[content-slug]/level-integration.md` | `design/narrative/[content-slug]-level-integration.md` |
| localization-lead | `production/narrative/[content-slug]/i18n-review.md` | stays in `production/` — a review, not a design document |

> **最終目的地不能隨便選。** `design/narrative/` 會被
> `/asset-spec`、`/design-review`、`/localize`、`/onboard`、
> `/project-stage-detect`、`/team-level` 和 `cd-narrative` director gate 讀取。
> 寫到其他地方的產物，這七個全都看不到。
>
> 上面 `[path]` 約定有個前提：路徑是*你*指定的。一個 skill 如果講了這個約定、卻沒指名具體的目的地，每次 run 都得自己發明一個，約定就形同虛設。`/team-qa` 的 Phase 4 基於同樣的理由，也有同樣的要求。

> **為什麼這不違反 Collaboration Protocol。** `CLAUDE.md` 要求 agent 在 Write/Edit 之前先問「May I write this to [filepath]?」。這裡 spawn 出來的 subagent **不問**就寫，這是刻意設計的有界例外，不是疏忽，跟 `consistency-check` 附加寫入 `active.md` 是同一個決定。這個例外只在三個條件**同時**成立時才有效：(1) 路徑是**你**在 prompt 裡指定的，所以使用者核准這個 phase 的時候，就等於核准了目的地；(2) 它是 `production/`、`docs/` 或 `tests/` 底下的新產物，絕對不是改既有的原始碼或設定；(3) 產生它的那個 phase，本身在 pipeline 往下走之前有 `AskUserQuestion` 把關。這三條以外，agent 一定要先問。**不要靠「每個 subagent 都問一次」來「修」這件事**：每個 agent 每個 phase 都跳一次提示，會讓 orchestrator 根本沒辦法用，這就是這個例外存在的原因。

只要 pipeline 允許，就把互相獨立的 agent 平行啟動（例如 Phase 2 的 agent 可以同時跑）。

## Pipeline

### Phase 1: Narrative Direction
委派給 **narrative-director**：
- 定義這份內容的敘事目的：它服務哪個劇情節拍？
- 找出涉及的角色、他們的動機，以及這跟整體弧線怎麼搭
- 設定情緒基調與節奏目標
- 指明任何背景設定的依賴項，或這份內容引入的新背景設定
- 輸出：附劇情需求的敘事 brief

### Phase 2: World Foundation (parallel)
平行委派：三個 `Agent` 呼叫要同時發出去，才去等任何結果：
- **world-builder**：為跟這份內容相關的陣營、地點和歷史建立或更新背景設定條目。對照既有的背景設定檢查有沒有矛盾。為新條目設定 canon 等級。
- **writer**：用聲音設定檔起草角色對白。確保每句都在 120 個字元以內、變數用具名佔位符、而且已為在地化做好準備。
- **art-director**：為這份內容中出現的關鍵角色定義角色視覺設計方向（剪影、視覺原型、辨識特徵）。為每個關鍵空間指定環境視覺敘事的元素（道具構成、光照筆記、空間安排）。為任何過場或腳本化的段落定義色調盤與電影化方向。

**Phase gate：** 收齊 Phase 2 的三份產出，然後在開始 Phase 3 之前套用「決策點」的規則：呈現背景設定、對白和視覺方向的摘要，在 `collaborative` 模式下透過 `AskUserQuestion` 取得核准（`guided` 下自動往下走，除非某個結果是 BLOCKED；`autonomous` 下用 `log_decision` 記錄）。

### Phase 3: Level Narrative Integration
委派給 **level-designer**：
- 檢視敘事 brief 與背景設定基礎
- 設計關卡裡的環境敘事元素
- 放置敘事觸發點、對白區域與發現點
- 確保節奏同時服務遊戲玩法和劇情

### Phase 4: Review and Consistency
透過 `Agent` spawn `narrative-director`，使用 gate **ND-CONSISTENCY**
（`.claude/docs/director-gates/nd-consistency.md`）。這個 phase 在每個 review mode 下都會跑：它是這條 pipeline 自己的審查，不是選用的 director gate。傳入那個 gate 列出的 context（Phase 2–3 的草稿路徑、敘事聖經或基調指南、世界規則、受影響的角色與陣營設定檔），這些已經在這裡點名了，所以不要在這個 session 裡讀 gate 檔案，然後再加上：
- 確認敘事節奏跟關卡設計一致
- 檢查所有謎團都有記錄在案的「真正答案」

依 verdict 處理：**APPROVE** → Phase 5。**CONCERNS** → 秀出列出的不一致處，並透過 `AskUserQuestion` 問要修改（重新 spawn 負責各項的 agent）還是接受。**REJECT** → 不要開始 Phase 5；矛盾要先修掉。**NOT ASSESSED [missing input]** → 指出缺的輸入，然後補上輸入再重跑這個 gate，或是把 `ND-CONSISTENCY: NOT ASSESSED — [input]` 帶進報告的一致性審查結果；絕對不能把它當成 APPROVE 來讀（`.claude/docs/director-gates.md`）。

**在 `team.size: individual` 下**，narrative-director 不在啟用名單內，所以不會 spawn 這個 gate。改由 writer 依同樣的標準檢查草稿，而且輸出要直接講明：「ND-CONSISTENCY not run — narrative-director is not
active at `team.size: individual`; consistency self-checked by writer.」自我檢查不是 gate 的 verdict，報告絕對不能把它當成 verdict 呈現。

### Phase 5: Polish (parallel)
平行委派：每個 `Agent` 呼叫都要發出去，才去等任何結果：
- **writer**：最後的自我審查：確認沒有任何一句超過對話框的限制、所有文字都用字串 key（不是原始字串）、佔位變數名稱一致
- **localization-lead**：驗證 i18n 合規性：檢查字串 key 的命名慣例、標出任何寫死格式、翻譯後會壞掉的字串、確認會變長的語言（德文／芬蘭文通常多 30%）有足夠的字元上限餘裕、確認文字中沒有需要地區專屬版本的文化假設
- **world-builder**：為所有新的背景設定條目定案 canon 等級

### Write the finals
所有草稿都在你手上。把上面表格裡的最終文件列出來，然後透過 `AskUserQuestion` 一次問清楚：「May I write these [N] narrative documents to
`design/narrative/`?」核准後再寫。`production/narrative/[content-slug]/` 底下的草稿保留下來當作工作紀錄。

## Error Recovery Protocol

**先驗證產物。** 如果回傳約定指定了路徑，在把這個 phase 當成做完之前，先確認該路徑真的存在：**指名的產物不在磁碟上，就是失敗的 phase，不管回覆讀起來多流暢。** agent 可能耗掉整個 phase，回你一段看似合理的開場白，卻什麼都沒寫；這既不是 BLOCKED、也不是錯誤、也不是「無法完成」，所以下面的觸發條件永遠不會被觸發。要叫它繼續，並指明沒達成的約定；context 通常還在。

如果任何 spawn 出去的 agent 回傳 BLOCKED、出錯或無法完成：**立刻揭露，不要越過它所阻擋的依賴項繼續往下走，而且一定要產出部分報告。** 被略過的 agent 那一節要留成一個有名字的缺口，絕對不准用你自己的內容填進去。完整流程：`.claude/docs/error-recovery-protocol.md`。

常見的阻擋原因：
- 輸入檔案缺失（找不到 story、GDD 不存在）→ 導向建立它的 skill
- ADR 狀態是 Proposed → 不要實作；決定之後，用 `/architecture-decision accept ADR-NNNN` 接受它
- 範圍太大 → 用 `/create-stories` 拆成兩個 story
- ADR 和 story 的指示互相衝突 → 把衝突揭露出來，不要猜

## File Write Protocol

Sub-agent 把它們的草稿寫到你指定的 `production/narrative/[content-slug]/` 路徑，依據的是上面「為什麼這不違反 Collaboration Protocol」所記載的**有界例外**：路徑是你指定的、產物是 `production/`、`docs/` 或 `tests/` 底下的新檔案、而且這個 phase 有 `AskUserQuestion` 把關。在這些範圍內，sub-agent **不會**每次寫入都問；範圍之外就一定要問。**唯一的例外是 `design/narrative/` 底下的最終文件**：所有草稿都在你手上，所以你在「Write the finals」那一次「May I write …?」之後，自己去寫。

## Output

一份摘要報告，涵蓋：敘事 brief 的狀態、建立／更新的背景設定條目、寫了多少對白、關卡敘事的整合點、一致性審查結果，以及任何未解決的矛盾，包括使用者選擇不修的在地化問題，要以字串 key 逐一列出。

Verdict: **COMPLETE** — narrative content delivered.

如果 ND-CONSISTENCY 回答 NOT ASSESSED、而且這件事已帶進報告（Phase 4），verdict 要講明，絕對不能只寫一個 COMPLETE：

Verdict: **COMPLETE — consistency NOT ASSESSED ([input])** — narrative content delivered; the consistency review could not run for want of [input].

如果 pipeline 因為有依賴項沒解決而停下來（例如使用者沒解決的背景設定矛盾或缺漏的前置條件）：

Verdict: **BLOCKED** — [reason]

## Next Steps

- 對敘事文件跑 `/design-review` 做一致性驗證。
- 對白定案之後，跑 `/localize extract` 把新字串抽出來交給翻譯。
- 跑 `/dev-story` 在引擎內實作對白觸發點和敘事事件。

---
name: team-narrative
description: "Orchestrate the narrative team — narrative-director, writer, world-builder, level-designer — for story, world lore, narrative-driven levels."
argument-hint: "[narrative content description] [--review full|lean|solo]"
user-invocable: true
allowed-tools: Read, Glob, Grep, Write, Edit, Agent, AskUserQuestion, TaskCreate, TaskGet, TaskList, TaskUpdate, Bash(bash "*/.claude/skills/team-narrative/../../hooks/yaml-helper.sh" resolve_config *)
model: sonnet
---
# Team Narrative

If no argument is provided, output usage guidance and exit without spawning any agents:
> Usage: `/team-narrative [narrative content description] [--review full|lean|solo]` — describe the story content, scene, or narrative area to work on (e.g., `boss encounter cutscene`, `faction intro dialogue`, `tutorial narrative`). Do not use `AskUserQuestion` here; output the guidance directly.

When this skill is invoked with an argument, orchestrate the narrative team through a structured pipeline.

**Decision Points:** At each phase transition, use `AskUserQuestion` to present
the user with the subagent's proposals as selectable options. Write the agent's
full analysis in conversation, then capture the decision with concise labels.
In `collaborative` mode, the user must approve before moving to the next phase.
In `guided` mode the pipeline advances automatically unless a phase is BLOCKED;
in `autonomous` mode it runs end to end, recording each phase outcome via
`log_decision`. Decisions in `automation_always_ask` categories
(`is_always_ask_category` helper) always prompt regardless of mode. See
`.claude/docs/automation-modes.md`.

## Phase 0: Resolve Config

!`bash "${CLAUDE_SKILL_DIR}/../../hooks/yaml-helper.sh" resolve_config --keys review_mode,automation,team.size`

Resolved above — use as-is; `--review` overrides `review_mode`. No block →
defaults in `.claude/docs/config-resolution.md`.

`review_mode` sets director-gate depth, and this pipeline has no director gate:
no phase below spawns CD-, TD-, PR- or AD-PHASE-GATE, at any `review_mode`. Its
phase gates are the pipeline's own decision points (defined under `team.size`
below), and the agents that work at them are team members, not director gates.

Phase 4's ND-CONSISTENCY is the exception in every mode: it is this pipeline's own
review, not an optional director gate, so `review_mode` never skips it (`team.size`
still decides who runs it — see Phase 4).

`automation` drives the Decision Points note above. See the Decision Points note above and
`.claude/docs/automation-modes.md` for how each mode changes pipeline behavior.

**`team.size`**: which agents are active (orthogonal to review_mode gate-depth and workflow docs).
- **`individual`** (default): `writer` only; `narrative-director` invoked only on an explicit pillar conflict. Other agents consulted via the writer, not spawned separately.
- **`small`**: `narrative-director` + `writer`, plus `localization-lead` and `world-builder` when `review_mode` is `full` (at `lean` and `solo` they are consulted through the writer).
- **`studio`**: all six Team Composition agents, whatever the `review_mode` (the full pipeline as documented).
A non-core agent needed at `individual` routes through the nearest active core agent with an informational note. **"Phase gate" means any phase that ends in an `AskUserQuestion` decision point this pipeline itself lists** — a transition under Decision Points above, or a **Gate** step written into the pipeline below — **whatever the `automation` mode.** `guided` and `autonomous` change how a gate is passed (it auto-advances, or is recorded with `log_decision`), not whether it is one, so bounded-exception condition (3) below holds at it in every mode. An agent restricted to "phase gates only" is spawned at those points and no others. This active-set scoping applies throughout the pipeline below: any phase that names an agent outside the active set routes through the nearest core agent rather than spawning it.

**Announce the active set before Phase 1 — never let the collapse be silent.**
Before spawning anything, state in one line which agents this run will actually
spawn, and which the pipeline below names but will **not** spawn at the resolved
`team.size`. For example:

> `Active set (team.size: <resolved>): <the agents listed for that size above>.`
> `Not spawned this run: <every other agent this pipeline names> — consulted`
> `through <nearest active core agent>. Raise team.size (or modes.rigor) to widen.`

Fill it from the `team.size` list directly above and the agents this file's own
pipeline names — not from an example. Both sets differ per orchestrator.

The pipeline below reads as a multi-agent fan-out and at the shipped default it
is one or two agents — `team-release` names ten and runs one, `team-narrative`
names six across five phases and runs `writer` alone. **The collapse is correct**:
`team.size` is rigor-fronted and the narrow default is the token lever.
Without saying so, a reader cannot
distinguish a correctly-collapsed run from a broken pipeline, and the per-agent
"routes through the nearest core agent with an informational note" rule above
fires at routing time and never states the shape of the run as a whole.

This is the same rule as the skipped-check reporting elsewhere in this file: **a constraint that is enforced but never surfaced is
indistinguishable, to the person reading the output, from one that was never
enforced.**

## Team Composition
- **narrative-director** — Story arcs, character design, dialogue strategy, narrative vision
- **writer** — Dialogue writing, lore entries, item descriptions, in-game text
- **world-builder** — World rules, faction design, history, geography, environmental storytelling
- **art-director** — Character visual design, environmental visual storytelling, cutscene/cinematic tone
- **level-designer** — Level layouts that serve the narrative, pacing, environmental storytelling beats
- **localization-lead** — Localization readiness — flags non-localizable strings, cultural assumptions, and i18n gaps

## How to Delegate

Use the `Agent` tool to spawn each team member as a subagent:
- `subagent_type: narrative-director` — Story arcs, character design, narrative vision
- `subagent_type: writer` — Dialogue writing, lore entries, in-game text
- `subagent_type: world-builder` — World rules, faction design, history, geography
- `subagent_type: art-director` — Character visual profiles, environmental visual storytelling, cinematic tone
- `subagent_type: level-designer` — Level layouts that serve the narrative, pacing
- `subagent_type: localization-lead` — Localization readiness — flags non-localizable strings, cultural assumptions, and i18n gaps

**Read the canon first.** Before Phase 1, read the shared inputs that exist — the narrative bible or tone guide, the character and faction files under `design/narrative/`, and the world rules. Never tell an agent to invent canon: if none exists, say so and ask the user before any drafting.

**Brief each agent — do not dump context.** Read the shared inputs **once** and pass a distilled brief inline: the lines each agent actually needs, never a file path for a document you have already read (an agent handed a path re-reads the whole file). Pass a path only for a document you have not read and only that agent needs.

**End every agent prompt with a return contract:** "Write your full output to `[path]` — that named path is your write authorisation under the bounded exception below, so write it without a separate approval prompt. Return **only** (1) the path written, (2) a ≤5-bullet summary of decisions, (3) any BLOCKED/CONCERNS items, one line each. Do not restate the documents you read." Without it, an agent returns everything it read back into this session.

**Substitute a real path for `[path]`.** Agents write drafts under
`production/narrative/[content-slug]/` — inside the bounded exception below. The
finished documents belong in `design/narrative/`, which the exception does **not**
cover, so **you** write those after one approval (see "Write the finals"):

| Agent | Drafts to (`[path]`) | Final document (you write it) |
|---|---|---|
| narrative-director | `production/narrative/[content-slug]/brief.md` | `design/narrative/[content-slug]-brief.md` |
| world-builder | `production/narrative/[content-slug]/lore-[topic].md` | `design/narrative/lore/[topic].md` |
| writer | `production/narrative/[content-slug]/dialogue-[scene-slug].md` | `design/narrative/dialogue/[scene-slug].md` |
| art-director | `production/narrative/[content-slug]/visual-direction.md` | `design/narrative/[content-slug]-visual-direction.md` |
| level-designer | `production/narrative/[content-slug]/level-integration.md` | `design/narrative/[content-slug]-level-integration.md` |
| localization-lead | `production/narrative/[content-slug]/i18n-review.md` | stays in `production/` — a review, not a design document |

> **The final destination is not a free choice.** `design/narrative/` is read by
> `/asset-spec`, `/design-review`, `/localize`, `/onboard`,
> `/project-stage-detect`, `/team-level` and the `cd-narrative` director gate.
> An artifact written anywhere else is invisible to all seven.
>
> The `[path]` contract above has a precondition: the path is one *you* named. A
> skill that states the contract but names no concrete destination leaves every
> run to invent one, which defeats it. `/team-qa` Phase 4 carries the same
> requirement for the same reason.

> **Why this does not violate the Collaboration Protocol.** `CLAUDE.md` requires an agent to ask "May I write this to [filepath]?" before Write/Edit. A subagent spawned here writes **without** asking, and that is a deliberate, bounded exception rather than an oversight — the same call already made for `consistency-check` appending to `active.md`. The exception holds only when all three are true: (1) the path is one **you** named in the prompt, so the user approved the destination when they approved the phase; (2) it is a new artifact under `production/`, `docs/` or `tests/`, never an edit to existing source or config; (3) the phase that produced it is itself gated by an `AskUserQuestion` before the pipeline advances. Outside those three, the agent must ask. **Do not "fix" this by asking per subagent** — a prompt per agent per phase makes an orchestrator unusable, which is why the exception exists.

Launch independent agents in parallel where the pipeline allows it (e.g., Phase 2 agents can run simultaneously).

## Pipeline

### Phase 1: Narrative Direction
Delegate to **narrative-director**:
- Define the narrative purpose of this content: what story beat does it serve?
- Identify characters involved, their motivations, and how this fits the overall arc
- Set the emotional tone and pacing targets
- Specify any lore dependencies or new lore this introduces
- Output: narrative brief with story requirements

### Phase 2: World Foundation (parallel)
Delegate in parallel — issue all three `Agent` calls simultaneously before waiting for any result:
- **world-builder**: Create or update lore entries for factions, locations, and history relevant to this content. Cross-reference against existing lore for contradictions. Set canon level for new entries.
- **writer**: Draft character dialogue using voice profiles. Ensure all lines are under 120 characters, use named placeholders for variables, and are localization-ready.
- **art-director**: Define character visual design direction for key characters appearing in this content (silhouette, visual archetype, distinguishing features). Specify environmental visual storytelling elements for each key space (prop composition, lighting notes, spatial arrangement). Define tone palette and cinematic direction for any cutscenes or scripted sequences.

**Phase gate:** collect all three Phase 2 outputs, then apply the Decision Points
rule before starting Phase 3 — present the lore, dialogue, and visual direction
summaries and capture approval via `AskUserQuestion` in `collaborative` mode
(auto-advance in `guided` unless a result is BLOCKED; `log_decision` in
`autonomous`).

### Phase 3: Level Narrative Integration
Delegate to **level-designer**:
- Review the narrative brief and lore foundation
- Design environmental storytelling elements in the level
- Place narrative triggers, dialogue zones, and discovery points
- Ensure pacing serves both gameplay and story

### Phase 4: Review and Consistency
Spawn `narrative-director` via `Agent` using gate **ND-CONSISTENCY**
(`.claude/docs/director-gates/nd-consistency.md`). This phase runs in every review
mode — it is this pipeline's own review, not an optional director gate. Pass the
context that gate lists (the Phase 2–3 draft paths, the narrative bible or tone
guide, the world rules, the affected character and faction profiles) — named
here, so do not read the gate file in this session — and add:
- Confirm narrative pacing aligns with level design
- Check that all mysteries have documented "true answers"

Act on the verdict: **APPROVE** → Phase 5. **CONCERNS** → show the listed
inconsistencies and ask via `AskUserQuestion` whether to revise them (re-spawn the
agent that owns each) or accept them. **REJECT** → do not start Phase 5; the
contradictions are fixed first. **NOT ASSESSED [missing input]** → name the
missing input, then supply it and re-run the gate, or carry
`ND-CONSISTENCY: NOT ASSESSED — [input]` into the report's consistency review
results; it is never read as APPROVE (`.claude/docs/director-gates.md`).

**At `team.size: individual`** narrative-director is not in the active set, so the
gate is not spawned. The writer checks the drafts against the same criteria, and
the output says so by name: "ND-CONSISTENCY not run — narrative-director is not
active at `team.size: individual`; consistency self-checked by writer." A
self-check is not the gate's verdict, and the report must not present it as one.

### Phase 5: Polish (parallel)
Delegate in parallel — issue every `Agent` call before waiting for any result:
- **writer**: Final self-review — verify no line exceeds dialogue box constraints, all text uses string keys (not raw strings), placeholder variable names are consistent
- **localization-lead**: Validate i18n compliance — check string key naming conventions, flag any strings with hardcoded formatting that won't survive translation, verify character limit headroom for languages that expand (German/Finnish typically +30%), confirm no cultural assumptions in text that would need locale-specific variants
- **world-builder**: Finalize canon levels for all new lore entries

### Write the finals
You hold every draft. List the final documents from the table above, then ask once
via `AskUserQuestion` — "May I write these [N] narrative documents to
`design/narrative/`?" — and write them on approval. The drafts under
`production/narrative/[content-slug]/` stay as the working record.

## Error Recovery Protocol

**First, verify the artifact.** If the return contract named a path, check the
path exists before treating the phase as done — **a named artifact that is not
on disk is a failed phase, however fluent the response reads.** An agent can
burn a full phase and return a plausible preamble having written nothing, which
is neither BLOCKED nor an error nor "cannot complete", so the trigger below
never fires. Resume it naming the unmet contract; the context is
usually still there.

If any spawned agent returns BLOCKED, errors, or cannot complete: **surface it
immediately, don't proceed past a dependency it blocks, and always produce a
partial report.** A skipped agent's section stays a named gap — never fill it with content of your own. Full procedure: `.claude/docs/error-recovery-protocol.md`.

Common blockers:
- Input file missing (story not found, GDD absent) → redirect to the skill that creates it
- ADR status is Proposed → do not implement; once it is decided, accept it with `/architecture-decision accept ADR-NNNN`
- Scope too large → split into two stories via `/create-stories`
- Conflicting instructions between ADR and story → surface the conflict, do not guess

## File Write Protocol

Sub-agents write their drafts to the `production/narrative/[content-slug]/` paths
you name, under the **bounded exception** documented above under "Why this does
not violate the Collaboration Protocol" — the path is one you named, the artifact
is new under `production/`, `docs/` or `tests/`, and the phase is gated by an
`AskUserQuestion`. A sub-agent does **not** prompt per write inside those bounds;
outside them it must ask. The **one exception is the final documents under
`design/narrative/`**: you hold every draft, so you write those yourself after
the single "May I write …?" in "Write the finals".

## Output

A summary report covering: narrative brief status, lore entries created/updated, dialogue lines written, level narrative integration points, consistency review results, and any unresolved contradictions — including any localization issue the user chose to leave unfixed, by string key.

Verdict: **COMPLETE** — narrative content delivered.

If ND-CONSISTENCY answered NOT ASSESSED and that was carried into the report
(Phase 4), the verdict says so — never a plain COMPLETE:

Verdict: **COMPLETE — consistency NOT ASSESSED ([input])** — narrative content delivered; the consistency review could not run for want of [input].

If the pipeline stops because a dependency is unresolved (e.g., lore contradiction or missing prerequisite not resolved by the user):

Verdict: **BLOCKED** — [reason]

## Next Steps

- Run `/design-review` on the narrative documents for consistency validation.
- Run `/localize extract` to extract new strings for translation after dialogue is finalized.
- Run `/dev-story` to implement dialogue triggers and narrative events in-engine.