mr-li-writer
Verified面向中文内容创作的编辑判断、研究、策划、写作与平台原生交付工作流。根据标题、素材、参考资料、核心想法或模糊灵感,先识别读者任务、材料条件和创新必要性,再按需完成语境扫描、可写性判断与方向校准,并区分发布平台与内容目标,进行事实检索、观点设计、平台原生正文写作、标题策划和对应交付。适用于公众号文章、知乎回答、小红书原生笔记、官网/网页、个人博客、行业解读、实用指南和基于素材的扩展研究写作。
scripts/lint_article.py
View on GitHub#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""Static checks for a Markdown article before delivery."""
import argparse
import re
import sys
from pathlib import Path
try:
from source_governance import (
compile_named_source_regex,
has_commercial_source_exposure,
has_restricted_source_mention,
source_role_names,
)
except ImportError:
from scripts.source_governance import (
compile_named_source_regex,
has_commercial_source_exposure,
has_restricted_source_mention,
source_role_names,
)
AI_PHRASES = (
"总而言之",
"综上所述",
"由此可见",
"值得注意的是",
"在当今时代",
"希望对你有所帮助",
"让我们一起",
"未来可期",
"拭目以待",
)
SEMANTIC_REVERSAL_PATTERNS = (
r"你以为.{2,80}(其实|实际|真正)",
r"看似.{2,80}(其实|实际|实则)",
r"表面上?.{2,80}(其实|实际|实则)",
r"不是.{2,80}而是",
r"并非.{2,80}而是",
r"不在于.{2,80}而在于",
r"(?:并不|不只|不再)是.{2,80}(?:而是|真正)",
)
NOMINALIZATION_PATTERNS = (
r"进行(?:了|一次|着)?.{0,12}(?:调整|优化|分析|讨论|梳理|复盘|探索|思考)",
r"实现(?:了)?.{0,16}(?:提升|增长|转变|突破)",
r"完成(?:了)?对.{0,20}的",
r"起到(?:了)?.{0,12}作用",
)
INSIGHT_ROAD_SIGNS = (
"更深一层",
"真正的问题是",
"更本质的是",
"还有一层",
"只说对了一半",
"答案恰恰相反",
)
CONJUNCTIONS = (
"因为",
"所以",
"但是",
"然而",
"同时",
"此外",
"因此",
"不仅",
"并且",
)
VIRTUAL_OPENING_PATTERNS = (
r"后台.*(有人|总有人|很多人|问)",
r"群里.*(有人|总有人|很多人|问)",
r"网上.*(有人|很多人|都在).*(问|讨论)",
r"评论区.*(经常|常常|总是|有人)",
r"有朋友问我",
r"最近.*很多人.*问",
r"每到.*(时候|季|季节|年底|年初|开学|报名).*(有人|很多人|问)",
)
GENERIC_OPENING_PATTERNS = (
(r"随着.{0,20}(发展|普及|变化|到来)", "开头疑似使用“随着……”宏大背景套话。"),
(r"在(当今|现代|如今).{0,12}(社会|时代|背景)", "开头疑似使用宏大背景开场。"),
(r"(你是否也曾|你有没有发现|有没有发现|不知道你有没有)", "开头疑似使用空泛设问。"),
(r"所谓.{1,20}(就是|指的是)", "开头疑似使用定义式套话。"),
(r"(朋友|同事|读者|学员)小[A-ZA-Z一二三四五六七八九十]", "开头疑似使用无真实素材支撑的假人物场景。"),
(r"最近.{0,20}(很火|刷屏|爆了|全网都在)", "开头疑似使用热点套话。"),
(r"每到.{0,12}(年底|年初|开学|报名|考试|毕业|换季|节假日)", "开头疑似使用泛季节/时间套话。"),
)
EVIDENCE_PHRASES = (
"数据显示",
"报告显示",
"研究表明",
"研究发现",
"专家指出",
"专家表示",
"调查显示",
"根据调查",
"根据研究",
"公开资料显示",
)
PROCESS_LEAK_PATTERNS = (
(r"(抓取|爬取|采集)(到|自|于|时间|结果|数据|页面|信息)?", "出现“抓取/爬取/采集”等后台采集动作词"),
(r"(检索|搜索|查询)(结果|显示|发现|到|出来)", "出现“检索/搜索结果”等生产过程表述"),
(r"(我|我们)?(通过|使用|借助).{0,12}(搜索引擎|爬虫|脚本|工具|模型|AI|大模型|提示词|prompt)", "暴露搜索、工具、模型或提示词过程"),
(r"(由|通过).{0,12}(AI|模型|大模型|系统).{0,12}(生成|整理|撰写|输出)", "暴露 AI/模型生成过程"),
(r"(本文|本篇|文章)?(?:为|是)?\s*(二创|二次创作|二创整理|基于原文改写|基于链接改写|基于素材改写)", "把内部改写/二创口径写进读者端正文"),
(r"(资料|数据|信息)(抓取|爬取|采集|清洗|抽取|汇总|整理)(时间|结果|口径)?", "把内部资料处理动作写进了正文"),
(r"(本次|这次).{0,8}(整理|检索|搜索|抓取|采集).{0,12}(发现|得到|结果)", "出现任务执行过程口吻"),
(r"(截至|截止)\s*\d{4}\s*年\s*\d{1,2}\s*月\s*(抓取|爬取|采集|检索)", "时间备注使用了后台动作词"),
(r"\d{4}\s*年\s*\d{1,2}\s*月\s*(抓取|爬取|采集|检索)", "时间备注使用了后台动作词"),
)
def configured_commercial_source_patterns():
path = Path(__file__).resolve().parents[1] / "references" / "commercial-source-terms.txt"
try:
terms = [
line.strip()
for line in path.read_text(encoding="utf-8").splitlines()
if line.strip() and not line.lstrip().startswith("#")
]
except FileNotFoundError:
print("[警告] 商业来源词表缺失: %s;将只使用通用商业来源模式。" % path, file=sys.stderr)
terms = []
if not terms:
return ()
regex = compile_named_source_regex(terms)
return (regex.pattern,) if regex else ()
OVER_LITERARY_PHRASES = (
"命运的齿轮",
"时代的洪流",
"灵魂深处",
"生命的底色",
"岁月长河",
"人间烟火",
"诗意地",
"在时光里",
"内心深处",
"精神旷野",
)
ABSTRACT_PHRASES = (
"意义",
"价值",
"本质",
"底层逻辑",
"时代",
"命运",
"灵魂",
"内核",
)
LOW_EVIDENCE_GENRES = {
"relationship-life",
"story-profile",
"personal-essay",
"platform-light",
"platform-native",
}
HIGH_RISK_MODES = {
"research-explainer",
}
XHS_MODULE_PATTERNS = (
r"适合(谁|人群|你)",
r"不适合(谁|人群|你)",
r"避坑",
r"判断标准",
r"可收藏|收藏清单|建议收藏",
)
XHS_LONGFORM_PATTERNS = (
r"本文将从",
r"接下来.{0,12}(展开|分析|讨论)",
r"从以下.{0,8}(方面|维度)",
r"随着.{0,20}(发展|普及|变化|到来)",
r"在(当今|现代|如今).{0,12}(社会|时代|背景)",
r"综上所述",
r"总而言之",
)
EMOJI_OR_SYMBOL_PATTERN = re.compile(
r"[\U0001F300-\U0001FAFF]|[✅⚠️📌✨🔥💡🌟👉❗️❤️⭐️]"
)
PERCENT_PATTERN = re.compile(r"(?<!\d)\d{1,3}(?:\.\d+)?%(?!\w)")
def parse_args():
parser = argparse.ArgumentParser(description="检查文章结构、引用和常见 AI 腔")
parser.add_argument("markdown", help="Markdown 文件")
parser.add_argument("--mode", default="research-explainer", help="内容模式")
parser.add_argument("--platform", default="", help="发布平台,如 公众号/知乎/小红书/官网/网页/个人博客")
parser.add_argument("--genre", default="", help="文章体裁,如 relationship-life/policy-industry/product-tool")
parser.add_argument(
"--evidence-density",
default="",
choices=("", "high", "medium", "low", "minimal", "高", "中", "低", "极低"),
help="证据密度: high/medium/low/minimal 或 高/中/低/极低",
)
parser.add_argument("--title", default="", help="可选标题,用于长度检查")
parser.add_argument("--task-state", default="", help="可选任务状态 JSON,用于按来源角色拦截不可显名机构")
parser.add_argument(
"--require-sources",
action="store_true",
help="要求文章包含参考资料和至少一个 URL",
)
parser.add_argument(
"--impact-check",
action="store_true",
help="检查高完成度增强是否走向过度文学化、抽象化或阅读负担过重",
)
parser.add_argument(
"--allow-commercial-source-names",
action="store_true",
help="文章本身在评测/介绍相关商业机构时,允许正文显名并要求说明利益关系",
)
parser.add_argument(
"--strict-delivery",
action="store_true",
help="正式交付模式:流程泄漏、商业资料背书、未经证实的第一人称和无来源百分比将阻断交付",
)
return parser.parse_args()
def han_count(text):
return len(re.findall(r"[\u4e00-\u9fff]", text))
def sentence_length_cv(text):
lengths = [
han_count(match.group())
for match in re.finditer(r"[^。!?!?\n]+[。!?!?]", text)
if han_count(match.group()) >= 4
]
if len(lengths) < 12:
return None
mean = sum(lengths) / len(lengths)
if not mean:
return None
variance = sum((length - mean) ** 2 for length in lengths) / len(lengths)
return (variance ** 0.5) / mean, len(lengths)
def main():
args = parse_args()
path = Path(args.markdown)
if not path.is_file():
print("[错误] 文件不存在: %s" % path, file=sys.stderr)
return 2
text = path.read_text(encoding="utf-8")
warnings = []
errors = []
platform = args.platform.strip().lower()
if not re.search(r"^#\s+\S+", text, re.M):
warnings.append("缺少一级标题,HTML 可能只能使用命令行传入的标题。")
body_start = re.sub(r"^#.*\n+", "", text.lstrip(), count=1)
opening_sample = body_start[:220]
if re.search(r"^(根据|数据显示|报告显示)", body_start.lstrip()):
warnings.append("开头直接进入来源或数据,检查是否缺少场景、问题或读者入口。")
for pattern in VIRTUAL_OPENING_PATTERNS:
if re.search(pattern, opening_sample):
warnings.append("开头疑似使用虚拟来源套话;请确认是否有真实读者提问,否则重写开头。")
break
for pattern, message in GENERIC_OPENING_PATTERNS:
if re.search(pattern, opening_sample):
warnings.append(message)
break
for phrase in AI_PHRASES:
count = text.count(phrase)
if count:
warnings.append("发现疑似模板化表达“%s” %d 次。" % (phrase, count))
reversal_hits = []
for pattern in SEMANTIC_REVERSAL_PATTERNS:
reversal_hits.extend(re.findall(pattern, text, re.S))
if reversal_hits:
warnings.append(
"发现疑似翻案修辞动作 %d 处;检查是否先替读者虚构误解再宣布洞察。真实的认识变化可以保留,换皮套路应改为直接判断和依据。"
% len(reversal_hits)
)
nominalization_hits = sum(
len(re.findall(pattern, text)) for pattern in NOMINALIZATION_PATTERNS
)
if nominalization_hits >= 2:
warnings.append(
"发现动词名词化表达 %d 处;检查能否还原成谁做了什么、改变了什么。"
% nominalization_hits
)
road_sign_hits = sum(text.count(phrase) for phrase in INSIGHT_ROAD_SIGNS)
if road_sign_hits >= 2:
warnings.append(
"洞察路标出现 %d 次;不要靠“更深一层/真正的问题”给段落排队,让材料和因果承担推进。"
% road_sign_hits
)
total_han = han_count(text)
conjunction_hits = sum(text.count(word) for word in CONJUNCTIONS)
if total_han >= 600 and conjunction_hits * 1000 / total_han > 9:
warnings.append(
"连词密度偏高;检查是否每段都靠因为/所以/然而等路标连接,能由语序和事理自然衔接的可删减。"
)
cv_result = sentence_length_cv(text)
if cv_result and cv_result[0] < 0.38:
warnings.append(
"全文 %d 个句子的长度过于接近;检查是否形成统一节拍。只在内容需要时调整长短,不要机械打散。"
% cv_result[1]
)
process_leaks = []
for pattern, message in PROCESS_LEAK_PATTERNS:
matches = re.findall(pattern, text, re.I)
if matches:
process_leaks.append(message)
if process_leaks:
message = (
"正文疑似暴露内容生产/资料处理过程:%s。请改为读者可接受的来源口径,如“信息截至 YYYY-MM-DD”“据官网当前页面”“公开资料显示”,不要出现抓取、爬取、采集、检索结果、AI 生成、提示词等后台动作。"
% ";".join(dict.fromkeys(process_leaks))
)
warnings.append(message)
if args.strict_delivery:
errors.append("严格交付门禁:%s" % message)
configured_patterns = configured_commercial_source_patterns()
configured_regex = re.compile(configured_patterns[0], re.I) if configured_patterns else None
governed_names, official_names = source_role_names(args.task_state)
governed_name_hit = has_restricted_source_mention(text, governed_names)
if not args.allow_commercial_source_names and (
governed_name_hit or has_commercial_source_exposure(text, configured_regex, official_names)
):
message = (
"正文疑似把商业相关第三方机构写成资料背书。硬信息应使用与该事实直接对应的官方来源;第三方仅作辅助核对时,请匿名改为“据相关第三方机构公开信息/公开汇总”,并在未获官方确认时明确“尚待官方确认”,不要在正文显名宣传。"
)
warnings.append(message)
if args.strict_delivery:
errors.append("严格交付门禁:%s" % message)
if re.search(r"我亲自|我亲身|我翻了[一二三四五六七八九十0-9]+份|我采访过", text):
message = "发现可能未经证实的第一人称经历或采访表达,请核对来源。"
warnings.append(message)
if args.strict_delivery:
errors.append("严格交付门禁:%s" % message)
if args.title:
title_length = len(args.title.strip())
if title_length < 8:
warnings.append("标题少于 8 个字符,检查主题识别度。")
if title_length > 60:
warnings.append("标题超过 60 个字符,检查搜索结果和移动端展示。")
if args.require_sources:
if not re.search(r"^##\s+(参考资料|References)\s*$", text, re.M):
errors.append("缺少“## 参考资料”或“## References”章节。")
if not re.search(r"https?://\S+", text):
errors.append("参考资料章节中没有可验证的 URL。")
if PERCENT_PATTERN.search(text) and not re.search(r"https?://\S+", text):
message = "正文包含百分比数据,但未发现 URL;请补充来源或删除数字。"
warnings.append(message)
if args.strict_delivery:
errors.append("严格交付门禁:%s" % message)
evidence_hits = sum(text.count(phrase) for phrase in EVIDENCE_PHRASES)
density = args.evidence_density
is_low_density = density in {"low", "minimal", "低", "极低"}
if args.genre in LOW_EVIDENCE_GENRES or is_low_density:
if evidence_hits >= 3:
warnings.append(
"低证据体裁/密度中出现数据、研究或专家话术 %d 次;检查是否喧宾夺主。"
% evidence_hits
)
if len(PERCENT_PATTERN.findall(text)) >= 2:
warnings.append("低证据文章包含多个百分比数字;检查是否把生活/关系内容写成报告。")
if args.require_sources and not re.search(r"^##\s+(参考资料|References)\s*$", text, re.M):
warnings.append("低证据文章未列参考资料可以接受,但请确认正文没有硬事实承诺。")
if args.mode in HIGH_RISK_MODES and density in {"", "low", "minimal", "低", "极低"}:
warnings.append("研究解释类文章使用低证据密度;请确认主题不是政策、价格、医疗、法律、金融等高风险硬信息。")
if args.impact_check:
literary_hits = sum(text.count(phrase) for phrase in OVER_LITERARY_PHRASES)
if literary_hits >= 2:
warnings.append("发现过度文学化表达 %d 次;检查是否影响平台读者理解和转述。" % literary_hits)
abstract_hits = sum(text.count(phrase) for phrase in ABSTRACT_PHRASES)
concrete_markers = len(re.findall(r"(早上|晚上|那天|一次|门口|桌上|手机|消息|说|问|看见|走|坐|放下|打开)", text))
if abstract_hits >= 10 and concrete_markers < 3:
warnings.append("抽象词较多但具体场景/动作较少;检查内容是否为了显得深刻而变飘。")
paragraphs = [p.strip() for p in re.split(r"\n\s*\n", text) if p.strip()]
long_paragraphs = [p for p in paragraphs if len(re.sub(r"\s+", "", p)) > 260]
if len(long_paragraphs) >= 3:
warnings.append("长段落较多;检查平台阅读是否吃力。")
# 判断可以由事实、动作和完整推理自然落下,不要求出现“真正/核心”等提示词。
is_xhs = (
"小红书" in args.platform
or "xiaohongshu" in platform
or "xhs" in platform
or args.mode == "xiaohongshu-note"
)
if is_xhs:
heading_match = re.search(r"^#\s+(.+)$", text, re.M)
if heading_match:
title = heading_match.group(1).strip()
if len(title) > 32:
warnings.append("小红书标题偏长;检查是否仍是 SEO/公众号长标题。")
else:
warnings.append("小红书笔记缺少一级标题。")
emoji_hits = len(EMOJI_OR_SYMBOL_PATTERN.findall(text))
if emoji_hits < 3:
warnings.append("小红书笔记少于 3 个 emoji 或符号化段落提示。")
if emoji_hits > 18:
warnings.append("小红书笔记 emoji/符号较多;检查是否满屏表情影响阅读。")
tags = re.findall(r"(?<!\S)#([A-Za-z0-9_\-\u4e00-\u9fff]+)", text)
if len(tags) < 6 or len(tags) > 10:
warnings.append("小红书标签数量应为 6-10 个,当前为 %d 个。" % len(tags))
module_hits = sum(1 for pattern in XHS_MODULE_PATTERNS if re.search(pattern, text))
if module_hits < 2:
warnings.append("小红书笔记缺少适合谁/不适合谁/避坑/判断标准/可收藏清单中的至少两个模块。")
paragraphs = [p.strip() for p in re.split(r"\n\s*\n", text) if p.strip()]
body_paragraphs = [p for p in paragraphs if not p.startswith("#") and not p.startswith("标签")]
long_xhs_paragraphs = [
p for p in body_paragraphs
if len(re.sub(r"\s+", "", p)) > 120 and not re.search(r"^[-*]\s+", p)
]
if len(long_xhs_paragraphs) >= 2:
warnings.append("小红书笔记存在多个过长段落;检查是否仍是公众号长文节奏。")
section_like_headings = len(re.findall(r"^#{2,3}\s+", text, re.M))
if section_like_headings >= 3:
warnings.append("小红书笔记出现较多长文小标题层级;检查是否像公众号/知乎结构。")
for pattern in XHS_LONGFORM_PATTERNS:
if re.search(pattern, text):
warnings.append("小红书笔记疑似出现公众号腔、长文铺垫或报告腔。")
break
if len(text) > 3200:
warnings.append("小红书笔记全文较长;检查是否只是把长文缩短而非原生笔记。")
if not re.search(r"标签\s*\n", text):
warnings.append("小红书笔记末尾建议使用“标签”区域集中放置话题标签。")
for warning in warnings:
print("[警告] %s" % warning)
for error in errors:
print("[错误] %s" % error)
if errors:
return 1
suffix = args.mode
if args.platform:
suffix += "/%s" % args.platform
if args.genre:
suffix += "/%s" % args.genre
if args.evidence_density:
suffix += "/%s" % args.evidence_density
print("[通过] %s (%s)" % (path, suffix))
return 0
if __name__ == "__main__":
sys.exit(main())
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 |