jianshuo/claude-skills / wangjianshuo-perspective
wangjianshuo-perspective
Unverified王建硕的思维框架与表达方式。基于 7 份深度调研(约 100 万词英文博客 + 约 109 万字中文博客, 2002–2022,全部一手)提炼出 7 个核心心智模型、10 条决策启发式和完整的双语表达 DNA。 用途:以王建硕的身份和声音写作、回应、思考——他平实、诚恳、好奇,爱用家常比喻和自造词, 在具体和抽象之间反复架梯子,从不写自己没亲身验证过的东西。 当用户提到「用王建硕的视角」「王建硕会怎么看」「王建硕模式」「像王建硕一样写」 「Jian Shuo Wang perspective」「切换到王建硕」时使用。 即使用户只是说「帮我用王建硕的角度想想」「如果是王建硕会怎么写这篇」也应触发。 一旦激活,本次对话后续所有回应都保持王建硕身份,直到用户明确说「退出」—— 不需要每轮重新点名。 不适用:用户要的是关于王建硕本人的客观介绍 / 事实查询(「王建硕是谁」「他哪年创业的」), 那种问题正常回答即可,不要进入角色扮演。
scripts/expression_stats.py
View on GitHub#!/usr/bin/env python3
"""Expression-DNA stats across the blog corpus + a stratified sample file.
Usage: python3 expression_stats.py [parsed_corpus_dir]
parsed_corpus_dir contains corpus_*.txt produced by parse_wxr.py
(defaults to ../references/sources/articles/parsed relative to this script)
"""
import re, os, glob, sys
from collections import Counter
_DEFAULT = os.path.join(os.path.dirname(__file__),
"..", "references", "sources", "articles", "parsed")
PARSED = sys.argv[1] if len(sys.argv) > 1 else os.path.normpath(_DEFAULT)
# Reassemble posts from corpus files
posts = []
for fn in sorted(glob.glob(os.path.join(PARSED, 'corpus_2*.txt'))):
txt = open(fn).read()
chunks = re.split(r'\n\n===== (.*?) =====\n\n', txt)
# chunks: ['', header, body, header, body, ...]
for i in range(1, len(chunks) - 1, 2):
header = chunks[i]
body = chunks[i + 1]
date = header.split(' | ')[0]
title = header.split(' | ')[1] if ' | ' in header else ''
posts.append({'date': date, 'title': title, 'body': body.strip()})
posts.sort(key=lambda p: p['date'])
print(f"Reassembled {len(posts)} posts")
allbody = "\n".join(p['body'] for p in posts)
words = re.findall(r"[a-zA-Z']+", allbody.lower())
print(f"Total word tokens: {len(words):,}")
STOP = set("""the a an and or but if then of to in on at for with as by is are was were be
been being it its this that these those i you he she we they my your his her our their
me him us them do does did have has had not no so very just can will would could should
may might from out up down about into over also there here what when where which who how
all one two more most some any no out s t re ll ve m d""".split())
content = [w for w in words if w not in STOP and len(w) > 2]
print("\n=== TOP 80 CONTENT WORDS ===")
for w, c in Counter(content).most_common(80):
print(f" {w}: {c}")
# Sentence length
sents = re.split(r'(?<=[.!?])\s+', allbody)
sents = [s for s in sents if 3 < len(s.split()) < 80]
avglen = sum(len(s.split()) for s in sents) / max(len(sents), 1)
short = sum(1 for s in sents if len(s.split()) <= 8)
print(f"\n=== SENTENCES ===")
print(f" count: {len(sents):,} avg words/sentence: {avglen:.1f}")
print(f" short (<=8 words): {short:,} ({100*short/len(sents):.0f}%)")
# Post openings (first sentence of each post)
print("\n=== 40 POST OPENINGS (sampled every Nth) ===")
step = max(1, len(posts) // 40)
for p in posts[::step]:
first = re.split(r'(?<=[.!?])\s+', p['body'])[0][:120]
print(f" [{p['date'][:10]}] {first}")
# Signature phrases
print("\n=== SIGNATURE PHRASE COUNTS ===")
for ph in ["I think", "I believe", "I don't know", "I am not sure", "I guess",
"in China", "in Shanghai", "for example", "as I", "to be honest",
"the fact that", "I would say", "interesting", "amazing", "I feel",
"my point is", "anyway", "P.S.", "Update:", "Wow", "lol",
"I am very", "very interesting", "It is interesting", "I love"]:
print(f" '{ph}': {len(re.findall(re.escape(ph), allbody, re.I))}")
# Stratified sample: ~3 posts per year, prefer mid-length
by_year = {}
for p in posts:
by_year.setdefault(p['date'][:4], []).append(p)
sample = []
for y in sorted(by_year):
ps = sorted(by_year[y], key=lambda x: abs(len(x['body'].split()) - 350))
sample.extend(ps[:3])
with open(os.path.join(PARSED, 'sample_stratified.txt'), 'w') as f:
for p in sample:
f.write(f"\n\n===== {p['date']} | {p['title']} =====\n\n{p['body']}\n")
print(f"\nWrote sample_stratified.txt: {len(sample)} posts")
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: 196k | 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: 196k | 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: 196k | Updated | Claude CodeCodex | Verified | Install |