Skip to content
Writing
Skill

/one-by-one-comparison-skill

逐一对比技能。将生成剧本与每个参考剧本逐一对比,按照剧本评价标准给出详细批判和改进建议。

BOOST
From plugin
novel-to-script-team
17925 skills17 agents
Install
$ npx -y skills add supreme-ultimate/novel-to-script-team --skill one-by-one-comparison-skill --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/one-by-one-comparison-skill

Context preview

The summary Claude sees to decide when to auto-load this skill.

逐一对比技能。将生成剧本与每个参考剧本逐一对比,按照剧本评价标准给出详细批判和改进建议。

SKILL.md

one-by-one-comparison-skill.SKILL.md
name: one-by-one-comparison-skill
description: 逐一对比技能。将生成剧本与每个参考剧本逐一对比,按照剧本评价标准给出详细批判和改进建议。

逐一对比技能

必读

1. `../../references/00-first-principles.md` 2. `../../references/03-script-writing-standard.md` 3. `../../references/04-review-gates.md`

技能说明

将生成剧本与检索到的Top 5参考剧本逐一对比,每次对比都按照统一的评价标准进行深度分析,给出具体的批判和改进建议。

评价标准(100分制)

1. 节奏控制(20分)

**评分细则**:

  • 场景数量(5分):与参考剧本相差±2个场景内得满分
  • 场景时长分配(5分):时长分布合理,有快有慢
  • 冲突密度(5分):冲突次数/总字数 ≥ 0.03
  • 节奏变化(5分):有明显的节奏起伏

**对比方法**:

# 统计生成剧本
gen_scenes = count_scenes(generated_script)
gen_conflicts = count_conflicts(generated_script)
gen_conflict_density = gen_conflicts / len(generated_script)

# 统计参考剧本
ref_scenes = count_scenes(reference_script)
ref_conflicts = count_conflicts(reference_script)
ref_conflict_density = ref_conflicts / len(reference_script)

# 评分
scene_score = 5 if abs(gen_scenes - ref_scenes) <= 2 else max(0, 5 - abs(gen_scenes - ref_scenes))
conflict_score = 5 if gen_conflict_density >= 0.03 else gen_conflict_density / 0.03 * 5

2. 对话风格(25分)

**评分细则**:

  • 对话比例(10分):70%+ 得满分,每低5%扣2分
  • 句长控制(8分):平均10-14字符得满分
  • 短句比例(4分):30-40%得满分
  • 对话自然度(3分):人物区分度、口语化程度

**对比方法**:

# 对话比例
gen_dialogue_ratio = calculate_dialogue_ratio(generated_script)
ref_dialogue_ratio = calculate_dialogue_ratio(reference_script)

dialogue_score = 10 if gen_dialogue_ratio >= 0.7 else gen_dialogue_ratio / 0.7 * 10

# 句长
gen_avg_length = calculate_avg_sentence_length(generated_script)
ref_avg_length = calculate_avg_sentence_length(reference_script)

if 10 <= gen_avg_length <= 14:
    length_score = 8
else:
    length_score = max(0, 8 - abs(gen_avg_length - 12) * 0.5)

3. 视觉化表达(20分)

**评分细则**:

  • 视觉标记密度(10分):3-5个/100字得满分
  • 表情描写(4分):冷笑、嗤笑、冷哼等
  • 眼神描写(3分):眼神一冷、眸光一沉等
  • 动作描写(3分):嘴角勾起、挑眉、转身等

**对比方法**:

visual_markers = ['冷笑', '嗤笑', '冷哼', '眼神一冷', '眸光一沉', '嘴角勾起', '挑眉', '转身']

gen_marker_density = count_markers(generated_script, visual_markers) / (len(generated_script) / 100)
ref_marker_density = count_markers(reference_script, visual_markers) / (len(reference_script) / 100)

if 3 <= gen_marker_density <= 5:
    marker_score = 10
else:
    marker_score = max(0, 10 - abs(gen_marker_density - 4) * 2)

4. 网文感(20分)

**评分细则**:

  • 网文感关键词密度(10分):1.5-2.5个/100字得满分
  • 情绪强化词(4分):冷笑、嗤笑、冷哼、冷声等
  • 态度词(3分):不屑、轻蔑、讥讽、嘲讽等
  • 打脸节奏(3分):打脸场景数量和质量

**对比方法**:

wanwen_keywords = {
    '情绪强化': ['冷笑', '嗤笑', '冷哼', '冷声', '冷冷地'],
    '态度词': ['不屑', '轻蔑', '讥讽', '嘲讽', '鄙夷'],
    '气场词': ['霸气', '强势', '凌厉', '锐利', '凛然'],
    '打脸词': ['啪啪打脸', '狠狠打脸', '当场打脸', '脸色一变']
}

gen_keyword_density = count_all_keywords(generated_script, wanwen_keywords) / (len(generated_script) / 100)
ref_keyword_density = count_all_keywords(reference_script, wanwen_keywords) / (len(reference_script) / 100)

if 1.5 <= gen_keyword_density <= 2.5:
    keyword_score = 10
else:
    keyword_score = max(0, 10 - abs(gen_keyword_density - 2) * 3)

5. 结构完整性(15分)

**评分细则**:

  • 开局钩子(3分):前10%是否有强冲突
  • 中段冲突(4分):中间60%是否持续有冲突
  • 高潮设计(3分):是否有明显高潮点
  • 反转设置(3分):是否有意外反转
  • 结尾悬念(2分):是否有强钩子

**对比方法**:

structure_elements = {
    'hook': check_opening_hook(script),  # True/False
    'development': check_continuous_conflict(script),
    'climax': check_climax(script),
    'twist': check_twist(script),
    'ending': check_ending_hook(script)
}

structure_score = sum([
    3 if structure_elements['hook'] else 0,
    4 if structure_elements['development'] else 0,
    3 if structure_elements['climax'] else 0,
    3 if structure_elements['twist'] else 0,
    2 if structure_elements['ending'] else 0
])

执行流程

Step 1: 加载剧本

# 加载生成剧本
with open(f'outputs/{剧本名}/scripts/ep{N}.md', 'r') as f:
    generated_script = f.read()

# 加载参考剧本列表(从检索结果或style-analysis报告中获取)
reference_files = [
    'knowledge/hit-scripts-md/参考剧本1.md',
    'knowledge/hit-scripts-md/参考剧本2.md',
    'knowledge/hit-scripts-md/参考剧本3.md',
    'knowledge/hit-scripts-md/参考剧本4.md',
    'knowledge/hit-scripts-md/参考剧本5.md'
]

reference_scripts = []
for ref_file in reference_files:
    with open(ref_file, 'r') as f:
        reference_scripts.append({
            'filename': os.path.basename(ref_file),
            'content': f.read()
        })

Step 2: 逐个对比

comparison_results = []

for i, ref in enumerate(reference_scripts, 1):
    print(f"\n{'='*80}")
    print(f"对比 {i}/5: 生成剧本 vs {ref['filename']}")
    print(f"{'='*80}\n")

    # 五大维度评分
    rhythm_score = compare_rhythm(generated_script, ref['content'])
    dialogue_score = compare_dialogue(generated_script, ref['content'])
    visual_score = compare_visual(generated_script, ref['content'])
    wanwen_score = compare_wanwen(generated_script, ref['content'])
    structure_score = compare_structure(generated_script, ref['content'])

    total_score = rhythm_score + dialogue_score + visual_score + wanwen_score + structure_score

    # 识别具体问题
    issues = identify_issues(generated_script, ref['content'])

    # 提取参考示例
    examples = extract_examples(ref['content'], issues)

    # 生成改进建议
    suggestions = generate_suggestions(issues, examples)

    comparison_results.append({
        'reference': ref['filename'],
        'scores': {
            'rhythm': rhythm_score,
            'dialogue': dialogue_score,
            'visual': visual_score,
            'wanwen': wanwen_score,
            'structure': structure_score,
            'total': total_score
        },
        'issues': issues,
        'examples': examples,
        'suggestions': suggestions
    })

Step 3: 综合分析

# 汇总共同问题
common_issues = find_common_issues(comparison_results)

# 计算平均得分
avg_scores = calculate_average_scores(comparison_results)

# 识别最弱维度
weakest_dimensions = identify_weakest_dimensions(avg_scores)

# 生成优先级建议
priority_suggestions = prioritize_suggestions(common_issues, weakest_dimensions)

Step 4: 生成报告

# 第{N}集逐一对比分析报告

## 一、综合评分

| 维度 | 平均得分 | 评级 |
|------|---------|------|
| 节奏控制 | 15/20 | B |
| 对话风格 | 18/25 | B+ |
| 视觉化表达 | 12/20 | C+ |
| 网文感 | 10/2
Read more
Ships withnovel-to-script-team

多 Agent、多 Skill 的小说改编影视流水线:把长篇小说、故事梗概或短剧素材,系统化转成改编分析、分集规划、剧本、审核报告、分镜方案、图片/视频提示词。

Get the whole plugin

Other skills on novel-to-script-team.