dongzhigang13305312738-art/paper-skills / complete-example
complete-example
Verified当用户明确要求"填充示例内容""生成示例""补充 LaTeX 示例"时使用。AI 增强版 LaTeX 示例智能生成器,实现 AI 与硬编码的有机融合:AI 做"语义理解"(分析章节主题、推理资源相关性、生成连贯叙述),硬编码做"结构保护"(格式验证、哈希校验、访问控制)。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.1 KB |
| config.yaml | config | 16.6 KB |
| references/README.md | markdown | 962 B |
| scripts/__init__.py | script | 503 B |
| scripts/advanced_usage.py | script | 5.4 KB |
| scripts/ai_content_generator.py | script | 19.6 KB |
| scripts/basic_usage.py | script | 1.3 KB |
| scripts/bibtex_parser.py | script | 5.6 KB |
| scripts/file_utils.py | script | 4.8 KB |
| scripts/format_guard.py | script | 15.5 KB |
| scripts/formula_generator.py | script | 2.1 KB |
| scripts/latex_parser.py | script | 6.2 KB |
| scripts/llm_client.py | script | 6.3 KB |
| scripts/placeholder_parser.py | script | 3.1 KB |
| scripts/resource_allocator.py | script | 10.3 KB |
| scripts/resource_scanner.py | script | 6.5 KB |
| scripts/security_manager.py | script | 17.4 KB |
| scripts/semantic_analyzer.py | script | 8.3 KB |
| scripts/skill_controller.py | script | 22.5 KB |
| scripts/table_generator.py | script | 5.8 KB |
| scripts/template_renderer.py | script | 994 B |
scripts/semantic_analyzer.py
View on GitHub"""
SemanticAnalyzer - AI 驱动语义分析器
🧠 AI:理解章节主题、推理资源相关性、评估内容质量
"""
from typing import List, Dict, Any, Optional
from dataclasses import dataclass
from pathlib import Path
import json
@dataclass
class SectionTheme:
"""章节主题分析结果"""
theme: str # 章节核心主题
key_concepts: List[str] # 关键概念列表
writing_style: str # 写作风格:学术/技术/说明
suggested_resources: List[str] # 建议的资源类型
tone: str # 语调:正式/通俗
target_audience: str # 目标读者:专家/评审/大众
@dataclass
class ResourceRelevance:
"""资源相关性评估结果"""
resource_path: str
relevance_score: float # 0-1 相关性分数
reason: str # AI 给出的理由
suggested_usage: str # 建议的使用方式
class SemanticAnalyzer:
"""AI 驱动的语义分析器"""
def __init__(self, llm_client, prompts: Optional[Dict[str, str]] = None):
"""
Args:
llm_client: LLM 客户端(Claude/OpenAI/本地模型)
prompts: 来自 config.yaml 的 prompts(可选,用于集中化管理)
"""
self.llm = llm_client
self.prompts = prompts or {}
def _safe_json(self, text: str) -> Dict[str, Any]:
"""尽量从 LLM 输出中解析 JSON;失败则返回空 dict。"""
try:
return json.loads(text)
except Exception:
# 兼容 LLM 返回包含说明文字的情况:截取第一个 JSON 对象
try:
start = text.find("{")
end = text.rfind("}")
if start != -1 and end != -1 and end > start:
return json.loads(text[start:end + 1])
except Exception:
pass
return {}
def analyze_section_theme(self, tex_content: str) -> SectionTheme:
"""
分析章节的核心主题和写作意图
Args:
tex_content: LaTeX 文件内容(前 2000 字符)
Returns:
SectionTheme: 结构化的主题分析结果
"""
tmpl = self.prompts.get("analyze_theme")
if tmpl:
prompt = tmpl.format(max_chars=2000, content=tex_content[:2000])
else:
prompt = f"""
你是一位专业的学术写作分析专家。请分析以下 LaTeX 章节的内容主题。
章节内容(前 2000 字符):
{tex_content[:2000]}
请返回 JSON 格式的分析结果:
{{
"theme": "章节的核心主题(一句话概括)",
"key_concepts": ["关键概念1", "关键概念2", "关键概念3"],
"writing_style": "学术/技术/说明/混合",
"suggested_resources": ["建议的图片类型", "建议的文献领域"],
"tone": "正式/半正式/通俗",
"target_audience": "评审专家/同行/学生/大众"
}}
"""
response = self.llm.complete(prompt, response_format="json", temperature=0.3)
data = self._safe_json(response)
# Robust defaults for mock/partial outputs.
return SectionTheme(
theme=str(data.get("theme") or "章节主题(未识别)"),
key_concepts=list(data.get("key_concepts") or ["关键概念A", "关键概念B", "关键概念C"]),
writing_style=str(data.get("writing_style") or "学术"),
suggested_resources=list(data.get("suggested_resources") or ["示意图", "相关文献"]),
tone=str(data.get("tone") or "正式"),
target_audience=str(data.get("target_audience") or "评审专家"),
)
def reason_resource_relevance(
self,
resource_info: 'ResourceInfo',
section_theme: SectionTheme
) -> ResourceRelevance:
"""
推理资源与章节的相关性
Args:
resource_info: 资源信息(图片/代码/文献)
section_theme: 章节主题
Returns:
ResourceRelevance: 相关性评估结果
"""
tmpl = self.prompts.get("reason_relevance")
if tmpl:
prompt = tmpl.format(
section_theme=section_theme.theme,
key_concepts=", ".join(section_theme.key_concepts),
writing_style=section_theme.writing_style,
resource_path=resource_info.path,
resource_type=resource_info.type,
metadata=resource_info.metadata,
)
else:
prompt = f"""
你是一位专业的学术写作顾问。请评估以下资源是否适合用于指定章节。
章节信息:
- 主题:{section_theme.theme}
- 关键概念:{', '.join(section_theme.key_concepts)}
- 写作风格:{section_theme.writing_style}
- 建议资源:{', '.join(section_theme.suggested_resources)}
资源信息:
- 路径:{resource_info.path}
- 类型:{resource_info.type}
- 元数据:{resource_info.metadata}
请返回 JSON 格式的评估结果:
{{
"relevance_score": 0.85,
"reason": "详细说明理由(100-200字)",
"suggested_usage": "建议如何使用这个资源"
}}
"""
response = self.llm.complete(prompt, response_format="json", temperature=0.3)
data = self._safe_json(response)
score = data.get("relevance_score")
try:
score = float(score)
except Exception:
score = 0.5
return ResourceRelevance(
resource_path=resource_info.path,
relevance_score=max(0.0, min(1.0, score)),
reason=str(data.get("reason") or "(未提供理由)"),
suggested_usage=str(data.get("suggested_usage") or "(未提供建议用法)"),
)
def generate_contextual_description(
self,
resource: 'ResourceInfo',
context: str,
usage_hint: str = None
) -> str:
"""
为资源生成符合上下文的描述文字
Args:
resource: 资源信息
context: 上下文内容(章节片段)
usage_hint: 使用提示(可选)
Returns:
str: 自然的描述文字(50-100 字)
"""
prompt = f"""
你是一位经验丰富的学术写作者。请为以下资源生成一段自然的描述文字。
上下文(章节片段):
{context[:500]}
资源信息:
- 文件名:{resource.filename}
- 类型:{resource.type}
- 建议用途:{usage_hint or '未指定'}
要求:
1. 生成 50-100 字的描述文字
2. 符合学术写作风格
3. 与上下文自然衔接
4. 包含适当的引入语(如"如图 X 所示"、"根据文献 X"等)
只返回描述文字,不要解释。
"""
return self.llm.complete(prompt, temperature=0.7)
def evaluate_content_quality(
self,
content: str,
section_theme: SectionTheme = None
) -> Dict[str, Any]:
"""
评估生成内容的质量
Args:
content: 生成的内容
section_theme: 章节主题(可选,用于对比)
Returns:
Dict: 质量评估报告
"""
prompt = f"""
你是一位严谨的学术写作评审专家。请评估以下示例内容的质量。
评估内容:
{content[:2000]}
章节主题(供参考):
{section_theme.theme if section_theme else '未指定'}
请返回 JSON 格式的评估报告:
{{
"coherence": 0.92, // 连贯性评分 0-1
"academic_tone": 0.88, // 学术风格评分 0-1
"resource_integration": "自然/生硬/无", // 资源整合评价
"strengths": ["优点1", "优点2"],
"weaknesses": ["不足1", "不足2"],
"suggestions": ["改进建议1", "改进建议2"],
"overall_score": 0.90 // 总体评分 0-1
}}
"""
response = self.llm.complete(prompt, response_format="json", temperature=0.3)
data = self._safe_json(response)
# Provide defaults so the pipeline can continue under MockLLMClient.
return {
"coherence": float(data.get("coherence") or 0.0),
"academic_tone": float(data.get("academic_tone") or 0.0),
"resource_integration": data.get("resource_integration") or "无",
"strengths": data.get("strengths") or [],
"weaknesses": data.get("weaknesses") or [],
"suggestions": data.get("suggestions") or [],
"overall_score": float(data.get("overall_score") or 0.0),
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |