topmindspace/topmind-writing-skills / topmind-x-article
topmind-x-article
Verified创作或整理 X Article 深度长文,支持论点推进、工具/插件介绍、产品实测评测与多源研究;按需导出可复制 HTML、纯文本、封面与发布清单。普通短帖用 topmind-briefs 或 topmind-viral-posts,纯格式化用 topmind-format。
scripts/info_design_hints.py
View on GitHub#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""info_design_hints.py — 信息设计提示(I1–I9,只提示,不判卷)。
python3 info_design_hints.py 稿子.md [--budget 2000] [--type essay] [--json]
判断方法见 references/information-design.md。提示的是「这里的内容形状可能更适合列表 / 表格 / 图」
「这一段可能在重复」,不是错误:永远不进 --strict,退出码只在读不到文件时为 2,其余为 0。
I1 连续纯文字块过多(≥4 段,中间没有图、表、列表、小标题)
I2 图、表、列表太少(长文每 800 字不到一个)
I3 一句里并列项多(「、」串起 3 项以上较长的项、分号串 3 段以上、连续几段用「第一 / 首先」开头),可以转列表
I4 连着几组「改前 / 改后」「A:/ B:」对照,可以转表格
I5 相邻段落重复表达(8 字以上相同片段),或一节首句复读小标题
I6 首段太长(超过 120 字),或首段没有交底(没有数字、以问句收住)
I7 某一节(##)超过 400 字却全是段落
I8 超出骨架里写的篇幅预算 30% 以上(读同目录 骨架.md 的「篇幅:约 N 字」,或 --budget)
I9 某一节(##)超过 400 字,却一句判断都没有(大概率 / 预计 / 意味着 / 我的判断 / 说白了…):素材罗列
frontmatter 写 `type: essay`(随笔、故事)时不提示 I1、I2、I7。
"""
import argparse
import json
import re
import sys
from difflib import SequenceMatcher
from pathlib import Path
CODES = {
"I1": "连续纯文字", "I2": "图表列表少", "I3": "并列项可转列表", "I4": "对照可转表格",
"I5": "重复表达", "I6": "首段没交底", "I7": "长节全是段落", "I8": "超篇幅预算",
"I9": "素材罗列",
}
ESSAY_EXEMPT = {"I1", "I2", "I7"}
RUN_MAX = 3 # 连续纯文字块的参考上限
CHARS_PER_ELEMENT = 800
LEAD_MAX = 120
SECTION_MIN = 400
REPEAT_MIN = 8
BUDGET_OVER = 0.30
# I9:判断句的标志词。一节里一个都没有,多半是在罗列素材(方法论、架构、对比类长文要有自己的判断)
JUDGE_RE = re.compile(r"大概率|预计|意味着|我的判断|我判断|我认为|我觉得|说白了|其实是|关键是|换句话说|所以|因此|应该|不如|值得|建议|更适合|看起来|多半|未必|恐怕|结论")
I9_EXEMPT_RE = re.compile(r"参考|来源|附录|延伸阅读|致谢|术语|名词")
_SYMBOLS_RE = re.compile(r"[\s*\-#>\[\]()`|]")
_URL_RE = re.compile(r"\]\([^)]*\)|https?://\S+")
LIST_RE = re.compile(r"^\s*(?:[-*+]\s|\d+[.)、)]\s?|[·•]\s?|[((]?\d+[))]\s?)")
ORDINAL_RE = re.compile(r"^(?:第[一二三四五六七八九十\d]+[,、,::]|首先|其次|再次|最后|其一|其二|其三)")
PAIR_RE = re.compile(r"^(?:改前|改后|原文|原句|原稿|改成|修改后|之前|之后|[ABAB][::]|方案\s*[ABAB])")
BOLD_HEAD_RE = re.compile(r"^\*\*[^*]{1,30}\*\*$") # 单独一行的加粗短语当小标题(路标)
CJK_RE = re.compile(r"[\u4e00-\u9fff]")
DIGIT_RE = re.compile(r"\d|[一二三四五六七八九十百千万亿两]+(?:个|次|倍|年|月|天|小时|分钟|字|篇|张|轮|成|%)")
def vlen(text: str) -> int:
return len(_SYMBOLS_RE.sub("", _URL_RE.sub("]", text)))
def split_frontmatter(lines):
meta, start = {}, 0
if lines and lines[0].strip() == "---":
for i in range(1, len(lines)):
if lines[i].strip() == "---":
for ln in lines[1:i]:
if ":" in ln:
k, v = ln.split(":", 1)
meta[k.strip()] = v.strip()
start = i + 1
break
return meta, start
def blocks_of(lines, start):
"""切成块:(类型, 起始行号, 文本)。类型:text / heading / image / list / table / quote / code / caption。"""
out, cur, cur_no, fence, comment = [], [], 0, False, False
def flush():
nonlocal cur
if cur:
out.append(classify(cur, cur_no))
cur = []
for i in range(start, len(lines)):
ln, no = lines[i], i + 1
s = ln.strip()
if comment:
if "-->" in s:
comment = False
continue
if s.startswith("<!--"):
flush()
if "-->" not in s:
comment = True
continue
if s.startswith(("```", "~~~")):
if not fence:
flush()
out.append(("code", no, ""))
fence = not fence
continue
if fence:
continue
if not s:
flush()
continue
if s.startswith("#") or BOLD_HEAD_RE.match(s):
flush()
out.append(("heading", no, s))
continue
if not cur:
cur_no = no
cur.append(s)
flush()
return out
def classify(buf, no):
first = buf[0]
text = "\n".join(buf)
if first.startswith("!["):
return ("image", no, text)
if first.startswith("|"):
return ("table", no, text)
if first.startswith(">"):
return ("quote", no, text)
if first.startswith(("来源:", "来源:", "图:", "注:")) and len(buf) == 1:
return ("caption", no, text)
if sum(1 for b in buf if LIST_RE.match(b)) >= max(2, len(buf) // 2 + 1):
return ("list", no, text)
if len(buf) >= 3 and sum(1 for b in buf if vlen(b) <= 40) == len(buf):
return ("list", no, text) # 短行分行写,读起来就是列表
return ("text", no, text)
ELEMENTS = {"image", "table", "list", "code", "quote"}
def units(text):
"""并列项长度:一个汉字算 1,一个英文词 / 数字算 1(「Mac、Windows、Linux」是短项,不提示)。"""
return len(CJK_RE.findall(text)) + len(re.findall(r"[A-Za-z0-9][A-Za-z0-9.+_-]*", text))
def _clauses(text):
return [c for c in re.split(r"[。!?!?\n]", text) if c.strip()]
def _strip_heading(h):
h = re.sub(r"^#+\s*", "", h)
h = re.sub(r"^[一二三四五六七八九十\d]+[、..]\s*", "", h)
return re.sub(r"[\s::,,??!!「」“”]", "", h)
def _norm(text):
return re.sub(r"[\s,。、;:!?,.;:!?「」“”‘’()()*`#>\-]", "", text)
def analyze(text, essay=False, budget=None):
lines = text.splitlines()
meta, start = split_frontmatter(lines)
if meta.get("type", "").strip().lower() == "essay":
essay = True
blocks = blocks_of(lines, start)
hits = []
body = [b for b in blocks if b[0] != "heading" or not b[2].startswith("# ")]
texts = [b for b in body if b[0] == "text"]
total = sum(vlen(b[2]) for b in body if b[0] in ("text", "list", "table", "quote", "caption"))
# I1 连续纯文字
run = []
for b in body + [("end", 0, "")]:
if b[0] == "text":
run.append(b)
continue
if b[0] == "caption":
continue
if len(run) > RUN_MAX:
hits.append((run[0][1], "I1", f"连续 {len(run)} 段纯文字(第 {run[0][1]}–{run[-1][1]} 行):中间能不能插一张图,或把并列 / 对照的内容转成列表、表格"))
run = []
# I2 图表列表少
elements = sum(1 for b in body if b[0] in ELEMENTS)
if total >= CHARS_PER_ELEMENT * 2 and total / max(1, elements) > CHARS_PER_ELEMENT:
hits.append((body[0][1] if body else 1, "I2", f"全文约 {total} 字,图 / 表 / 列表 {elements} 个:若有数据、步骤或对比可考虑换形式;纯文字也可成立,不设配图配额"))
# I3 并列项
ord_run = []
for b in body + [("end", 0, "")]:
if b[0] == "text" and ORDINAL_RE.match(b[2]):
ord_run.append(b)
else:
if len(ord_run) >= 2:
hits.append((ord_run[0][1], "I3", f"连续 {len(ord_run)} 段用「第一 / 首先」类开头:若是并列信息可用列表;有叙事作用时保留"))
ord_run = []
for b in texts:
done = False
for c in _clauses(b[2]):
semis = [x for x in re.split(r"[;;]", c) if vlen(x) >= 4]
if len(semis) >= 3:
hits.append((b[1], "I3", f"一句里用分号串了 {len(semis)} 段:可以拆成列表"))
done = True
break
for part in re.split(r"[,,::]", c):
items = part.split("、")
if len(items) >= 3 and min(units(x) for x in items[:-1]) >= 4 and units(items[-1]) >= 2:
hits.append((b[1], "I3", f"「{part[:20]}…」并列了 {len(items)} 项、每项都不短:可以转成列表"))
done = True
break
if done:
break
# I4 对照
pair_lines = []
for b in texts + [b for b in body if b[0] == "quote"]:
for k, ln in enumerate(b[2].split("\n")):
if PAIR_RE.match(re.sub(r"^[>\s*]+", "", ln)):
pair_lines.append(b[1] + k)
if len(pair_lines) >= 4:
hits.append((pair_lines[0], "I4", f"正文里有 {len(pair_lines)} 行「改前 / 改后」「A:/ B:」式对照:两组以上可以做成两列表格"))
# I5 重复:相邻(3 段内)段落有 8 字以上相同片段;一节首句复读小标题
norm = [(b[1], _norm(b[2])) for b in texts if not PAIR_RE.match(b[2])] # 对照行本来就相似,归 I4
seen = set()
for i in range(len(norm)):
for j in range(i + 1, min(i + 3, len(norm))):
a, c = norm[i][1], norm[j][1]
m = SequenceMatcher(None, a, c, autojunk=False).find_longest_match(0, len(a), 0, len(c))
frag = a[m.a:m.a + m.size]
if m.size >= REPEAT_MIN and len(CJK_RE.findall(frag)) >= 6 and norm[j][0] not in seen:
seen.add(norm[j][0])
hits.append((norm[j][0], "I5", f"和第 {norm[i][0]} 行重复「{frag[:16]}」:检查重复是否帮助强调、续读或解释,没有作用再合并"))
for k, b in enumerate(body):
if b[0] == "heading" and b[2].startswith("##"):
core = _strip_heading(b[2])
nxt = next((x for x in body[k + 1:] if x[0] != "caption"), None)
if nxt and nxt[0] == "text" and len(core) >= 4:
first = _norm(_clauses(nxt[2])[0]) if _clauses(nxt[2]) else ""
m = SequenceMatcher(None, core, first, autojunk=False).find_longest_match(0, len(core), 0, len(first))
if m.size >= max(4, int(len(core) * 0.8)):
hits.append((nxt[1], "I5", f"首句复读小标题「{core[:12]}」:判断复述是否帮助进入本节,必要时保留或给新信息"))
# I6 首段
lead = next((b for b in body if b[0] in ("text", "list", "table", "image")), None)
if lead and lead[0] == "text":
n = vlen(lead[2])
if n > LEAD_MAX:
hits.append((lead[1], "I6", f"首段 {n} 字:根据读者选择事实直给、故事或问题入口,不按字数自动改写"))
elif not DIGIT_RE.search(lead[2]) and re.search(r"[??]\s*$|[…]+\s*$", lead[2].strip()):
hits.append((lead[1], "I6", "首段以问句或省略号收住且未命中数字:看它是否形成有效问题或叙事入口,不据此判定缺结论"))
# I7 长节全是段落
sec, sec_no, sec_title = [], None, ""
for b in body + [("heading", 0, "## end")]:
if b[0] == "heading" and b[2].startswith("## "):
if sec_no and sec:
chars = sum(vlen(x[2]) for x in sec if x[0] == "text")
if chars > SECTION_MIN and not any(x[0] in ELEMENTS for x in sec):
hits.append((sec_no, "I7", f"这一节约 {chars} 字,全是段落:如果是比较或步骤,可考虑图、列表、表格或分节;连续论证与叙事可保留"))
all_chars = sum(vlen(str(x[2])) for x in sec)
if (all_chars > SECTION_MIN and not I9_EXEMPT_RE.search(sec_title)
and not any(JUDGE_RE.search(str(x[2])) for x in sec)):
hits.append((sec_no, "I9", f"这一节约 {all_chars} 字,未命中这份判断词表:按体裁看是否需要作者判断,事实汇编与教程不强加立场"))
sec, sec_no, sec_title = [], b[1], b[2]
elif sec_no:
sec.append(b)
# I8 篇幅预算
if budget and total > budget * (1 + BUDGET_OVER):
hits.append((body[0][1] if body else 1, "I8", f"全文约 {total} 字,骨架预算 {budget} 字,超出 {total / budget - 1:.0%}:复查预算是否仍适合当前任务,必要时调整预算或内容"))
if essay:
hits = [h for h in hits if h[1] not in ESSAY_EXEMPT]
return sorted(hits, key=lambda h: (h[0], h[1]))
def read_budget(md_path: Path):
sk = md_path.parent / "骨架.md"
if not sk.is_file():
return None
m = re.search(r"篇幅[::]\s*约?\s*(\d{3,6})\s*字", sk.read_text(encoding="utf-8", errors="replace"))
return int(m.group(1)) if m else None
def check_file(path, essay=False, budget=None):
p = Path(path)
text = p.read_text(encoding="utf-8-sig")
return analyze(text, essay=essay, budget=budget or read_budget(p))
def main(argv=None):
ap = argparse.ArgumentParser(description="信息设计提示 I1–I9(只提示,不进 --strict)")
ap.add_argument("files", nargs="+")
ap.add_argument("--budget", type=int, help="篇幅预算(字);缺省读同目录 骨架.md")
ap.add_argument("--type", choices=["essay"], help="随笔 / 故事:不提示 I1、I2、I7")
ap.add_argument("--json", action="store_true")
args = ap.parse_args(argv)
report = []
for f in args.files:
try:
hits = check_file(f, essay=args.type == "essay", budget=args.budget)
except (OSError, UnicodeDecodeError) as e:
print(f"错误:无法读取 {f}({e})", file=sys.stderr)
return 2
report.append({"file": f, "hints": [{"line": n, "code": c, "msg": m} for n, c, m in hits]})
if not args.json:
for n, c, m in hits:
print(f"{f}:{n}: [{c}] 信息设计:{m}")
if args.json:
print(json.dumps(report, ensure_ascii=False, indent=1))
else:
total = sum(len(r["hints"]) for r in report)
print(f"info_design_hints:{total} 条提示(只供判断,不是错误;方法见 information-design.md)")
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 |