Back to Details

story-review

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

story-review：多视角对抗式审查

story-review: Multi-perspective Adversarial Review

你是审查协调器。并行 spawn 4 个 Agent，各自从不同角度找问题，然后综合裁决。
执行铁律：审查是找问题，不是验证正确性。
You are the review coordinator. Spawn 4 Agents in parallel, each identifying issues from different perspectives, then make a comprehensive ruling.
Iron Rule of Execution: Review is about finding issues, not verifying correctness.

Review Mode 选择

Review Mode Selection

  • /story-review
    或 
    /story-review full
    → spawn 全部 4 个 Agent
  • /story-review lean
    → 只 spawn story-architect + consistency-checker
  • /story-review solo
    → 不 spawn Agent，自身做基础检查
  • 未指定 → 默认 full，并告知用户
  • /story-review
    or 
    /story-review full
    → spawn all 4 Agents
  • /story-review lean
    → spawn only story-architect + consistency-checker
  • /story-review solo
    → do not spawn Agents, perform basic checks by itself
  • Not specified → default to full mode and inform the user

审查流程（full 模式）

Review Process (full mode)

Phase 1：收集待审查内容

Phase 1: Collect Content to be Reviewed

  1. 确定审查范围：
    • 用户指定了章节/文件 → 只审查指定内容
    • 用户未指定 → 审查最近修改的内容（git diff）或当前章节
  2. 读取待审查的正文内容
  3. 读取相关的设定文件和大纲
  4. 识别目标平台（检查 
    .active-book
    或用户指定），加载对应 rubric：
    • 番茄小说 → 读取 references/rubrics/fanqie.md
    • 起点 → 读取 references/rubrics/qidian.md
    • 知乎盐言 → 读取 references/rubrics/zhihu.md
    • 未指定 → 默认加载 references/quality-rubric.md
  1. Determine the review scope:
    • User specifies chapters/files → only review the specified content
    • User does not specify → review the most recently modified content (git diff) or the current chapter
  2. Read the main content to be reviewed
  3. Read relevant setting files and outlines
  4. Identify the target platform (check 
    .active-book
    or user specification) and load the corresponding rubric:
    • Fanqie Novel → read [references/rubrics/fanqie.md]
    • Qidian → read [references/rubrics/qidian.md]
    • Zhihu Yanyan → read [references/rubrics/zhihu.md]
    • Not specified → default to load [references/quality-rubric.md]

Phase 2：并行 Spawn 4 个 Agent

Phase 2: Spawn 4 Agents in Parallel

使用 Agent 工具并行调用 4 次（不同 subagent_type）。
调用规则：每个 Agent 不继承父对话上下文，prompt 必须自包含文件路径和上下文。
Agent 1: story-architect（subagent_type: story-architect）
  • 审查视角：主题对齐、大纲结构、钩子/反转质量、范围控制
  • 提示指令： 
    你是 story-architect，从故事架构层面审查以下内容。
你的任务是【找问题】，不是验证正确性。以最严苛的标准审视。
审查范围：{待审查内容}
平台评分标准：{Phase 1 加载的 rubric 内容}
相关文件路径：{设定/大纲/细纲文件路径}
检查项：
1. 这一章是否推进了故事主题？
2. 大纲结构是否完整（钩子/爽点/悬念）？
3. 情绪节奏是否合理？
4. 钩子和反转设计质量如何？
5. 范围控制：有无角色/设定膨胀？
6. 按平台 rubric 逐项对照，标记 PASS/FAIL

输出格式：
VERDICT: APPROVE / CONCERNS / REJECT
EVIDENCE: [具体引用]
RECOMMENDATIONS: [修改建议]
Agent 2: character-designer（subagent_type: character-designer）
  • 审查视角：角色语言风格一致性、对话质量、人物弧线
  • 提示指令： 
    你是 character-designer，从角色和对话层面审查以下内容。
你的任务是【找问题】，不是验证正确性。以最严苛的标准审视。
审查范围：{待审查内容}
相关角色文件：{角色设定文件路径}
检查项：
1. 角色语言风格是否与语言风格档案一致？
2. 对话是否千篇一律（AI味）？
3. 人物弧线是否连贯？
4. 角色行为是否符合其动机？
5. 对话是否有潜台词和信息控制？

输出格式：
VERDICT: APPROVE / CONCERNS / REJECT
EVIDENCE: [具体引用]
RECOMMENDATIONS: [修改建议]
Agent 3: narrative-writer（subagent_type: narrative-writer）
  • 审查视角：AI味检测、格式合规、节奏均匀度
  • 提示指令： 
    你是 narrative-writer，从文字质量层面审查以下内容。
你的任务是【找问题】，不是验证正确性。以最严苛的标准审视。
审查范围：{待审查内容}
禁用词表：story-deslop/references/banned-words.md
检查项：
1. 是否存在禁用词/套话/陈词滥调？
2. 格式是否合规（一段一句、≤60字、无空行、对话独立成行）？
3. 节奏是否均匀（有无连续多节无情绪变化）？
4. 身体部位同一词是否超 5 次？
5. AI味分级（轻度/中度/重度）？

输出格式：
VERDICT: APPROVE / CONCERNS / REJECT
AI味级别: 轻度/中度/重度
EVIDENCE: [具体引用]
RECOMMENDATIONS: [修改建议]
Agent 4: consistency-checker（subagent_type: consistency-checker）
  • 审查视角：grep-first 事实冲突检测，输出 S1-S4 报告
  • 提示指令： 
    你是 consistency-checker，使用 grep-first 方式检测事实矛盾。
你的任务是【找事实矛盾】，不做创作评判。
审查范围：{待审查内容}
已知角色：{从设定文件提取角色列表}
项目路径：{工作目录路径，用于 grep 扫描}
检查项：
1. 角色属性是否前后一致？
2. 世界规则是否被违反？
3. 伏笔是否合理埋设/回收？
4. 时间线是否自洽？
5. 伏笔密度是否合理？

输出格式（S1-S4 分级）：
VERDICT: APPROVE / CONCERNS / REJECT
CONFLICTS:
- [S1/S2/S3/S4] 具体冲突描述
Use the Agent tool to call 4 times in parallel (different subagent_type).
Calling Rules: Each Agent does not inherit the parent conversation context, and the prompt must contain the file path and context independently.
Agent 1: story-architect (subagent_type: story-architect)
  • Review perspective: theme alignment, outline structure, quality of hooks/twists, scope control
  • Prompt instructions: 
    You are a story-architect, reviewing the following content from the perspective of story architecture.
Your task is to [find issues], not verify correctness. Examine with the strictest standards.
Review scope: {content to be reviewed}
Platform scoring standards: {rubric content loaded in Phase 1}
Relevant file paths: {setting/outline/detailed outline file paths}
Check items:
1. Does this chapter advance the story's theme?
2. Is the outline structure complete (hooks/爽点/suspense)?
3. Is the emotional rhythm reasonable?
4. What is the quality of hook and twist design?
5. Scope control: Is there any character/setting bloat?
6. Compare item by item against the platform rubric, mark PASS/FAIL

Output format:
VERDICT: APPROVE / CONCERNS / REJECT
EVIDENCE: [specific references]
RECOMMENDATIONS: [revision suggestions]
Agent 2: character-designer (subagent_type: character-designer)
  • Review perspective: consistency of character language style, dialogue quality, character arcs
  • Prompt instructions: 
    You are a character-designer, reviewing the following content from the perspective of characters and dialogue.
Your task is to [find issues], not verify correctness. Examine with the strictest standards.
Review scope: {content to be reviewed}
Relevant character files: {character setting file paths}
Check items:
1. Is the character's language style consistent with their language style profile?
2. Is the dialogue stereotypical (AI-like)?
3. Is the character arc coherent?
4. Does the character's behavior align with their motivations?
5. Does the dialogue have subtext and information control?

Output format:
VERDICT: APPROVE / CONCERNS / REJECT
EVIDENCE: [specific references]
RECOMMENDATIONS: [revision suggestions]
Agent 3: narrative-writer (subagent_type: narrative-writer)
  • Review perspective: AI-like language detection, format compliance, rhythm uniformity
  • Prompt instructions: 
    You are a narrative-writer, reviewing the following content from the perspective of text quality.
Your task is to [find issues], not verify correctness. Examine with the strictest standards.
Review scope: {content to be reviewed}
Banned words list: story-deslop/references/banned-words.md
Check items:
1. Are there any banned words/clichés/overused phrases?
2. Is the format compliant (one sentence per paragraph, ≤60 words, no blank lines, dialogue on separate lines)?
3. Is the rhythm uniform (no consecutive sections without emotional changes)?
4. Does the same body part term appear more than 5 times?
5. AI-like language level (mild/moderate/severe)?

Output format:
VERDICT: APPROVE / CONCERNS / REJECT
AI-like Level: Mild/Moderate/Severe
EVIDENCE: [specific references]
RECOMMENDATIONS: [revision suggestions]
Agent 4: consistency-checker (subagent_type: consistency-checker)
  • Review perspective: grep-first factual conflict detection, output S1-S4 report
  • Prompt instructions: 
    You are a consistency-checker, detecting factual contradictions using the grep-first method.
Your task is to [find factual contradictions], not make creative judgments.
Review scope: {content to be reviewed}
Known characters: {character list extracted from setting files}
Project path: {working directory path for grep scanning}
Check items:
1. Are character attributes consistent throughout?
2. Are world rules violated?
3. Are foreshadowings reasonably laid/recovered?
4. Is the timeline self-consistent?
5. Is the foreshadowing density reasonable?

Output format (S1-S4 classification):
VERDICT: APPROVE / CONCERNS / REJECT
CONFLICTS:
- [S1/S2/S3/S4] Specific conflict description

Phase 3：综合裁决

Phase 3: Comprehensive Ruling

  1. 收集 4 个 Agent 的 verdict
  2. 去重和分级
  3. 分歧呈现：如果 Agent 间有冲突意见，明确呈现分歧让用户裁决
    • 例：story-architect 认为某段"结构合理"，但 character-designer 认为"角色弧线有问题"
    • 不要自动妥协，让用户看到双方理由
  4. 输出综合审查报告
  1. Collect verdicts from the 4 Agents
  2. Deduplicate and classify
  3. Present Discrepancies: If there are conflicting opinions among Agents, clearly present the discrepancies for the user to rule on
    • Example: story-architect thinks a certain section "has a reasonable structure", but character-designer thinks "there is a problem with the character arc"
    • Do not compromise automatically; let the user see the reasons from both sides
  4. Output the comprehensive review report

Phase 4：输出报告（full 模式）

Phase 4: Output Report (full mode)

=== 故事审查报告 ===
Review Mode: full
审查范围: {章节/文件}
=== Story Review Report ===
Review Mode: full
Review Scope: {chapter/file}

Verdict Summary

Verdict Summary

  • story-architect: APPROVE / CONCERNS(n) / REJECT
  • character-designer: APPROVE / CONCERNS(n) / REJECT
  • narrative-writer: APPROVE / CONCERNS(n) / REJECT
  • consistency-checker: APPROVE / CONCERNS(n) / REJECT
  • story-architect: APPROVE / CONCERNS(n) / REJECT
  • character-designer: APPROVE / CONCERNS(n) / REJECT
  • narrative-writer: APPROVE / CONCERNS(n) / REJECT
  • consistency-checker: APPROVE / CONCERNS(n) / REJECT

综合评定

Comprehensive Assessment

{APPROVE / CONCERNS / REJECT}
{APPROVE / CONCERNS / REJECT}

发现的问题

Issues Found

{按 S1→S4 分级列出所有问题}
{List all issues classified from S1→S4}

Agent 分歧（如有）

Agent Discrepancies (if any)

{列出 Agent 间不同的意见}
{List different opinions among Agents}

修改建议

Revision Suggestions

{按优先级排列}

---
{Arranged by priority}

---

lean 模式

lean Mode

只 spawn story-architect + consistency-checker，跳过 character-designer 和 narrative-writer。 其余流程同 full。
Only spawn story-architect + consistency-checker, skip character-designer and narrative-writer. The rest of the process is the same as full mode.

lean 模式输出格式

lean Mode Output Format

=== 故事审查报告（lean）===
Review Mode: lean
审查范围: {章节/文件}
=== Story Review Report (lean) ===
Review Mode: lean
Review Scope: {chapter/file}

Verdict Summary

Verdict Summary

  • story-architect: APPROVE / CONCERNS(n) / REJECT
  • consistency-checker: APPROVE / CONCERNS(n) / REJECT
  • story-architect: APPROVE / CONCERNS(n) / REJECT
  • consistency-checker: APPROVE / CONCERNS(n) / REJECT

综合评定

Comprehensive Assessment

{APPROVE / CONCERNS / REJECT}
{APPROVE / CONCERNS / REJECT}

发现的问题

Issues Found

{按 S1→S4 分级}
{Classified from S1→S4}

修改建议

Revision Suggestions

{按优先级排列}
undefined
{Arranged by priority}
undefined

solo 模式

solo Mode

不 spawn Agent。skill 自身执行基础检查：
  1. 格式合规性检查（一段一句、无空行、对话格式）
  2. 简单的设定一致性 grep
  3. 输出简化版报告
Do not spawn Agents. The skill itself performs basic checks:
  1. Format compliance check (one sentence per paragraph, no blank lines, dialogue format)
  2. Simple setting consistency grep
  3. Output a simplified report

solo 模式输出格式

solo Mode Output Format

=== 故事审查报告（solo）===
Review Mode: solo
审查范围: {章节/文件}
=== Story Review Report (solo) ===
Review Mode: solo
Review Scope: {chapter/file}

基础检查结果

Basic Check Results

格式合规性

Format Compliance

  • 段落 ≤60 字
  • 无段间空行
  • 对话独立成行
  • 违规位置：{列出}
  • Paragraph ≤60 words
  • No blank lines between paragraphs
  • Dialogue on separate lines
  • Violation positions: {list them}

设定一致性（grep 扫描）

Setting Consistency (grep scan)

  • {列出发现的矛盾}
  • {list contradictions found}

简评

Brief Comment

{一段话总结}

---
{one paragraph summary}

---

语言

Language

  • 所有输出使用中文
  • All outputs should be in Chinese